Andrew O. Finley, Jeffrey W.

Doser, Vince Melfi

R Programming for Data

Sciences
Contents

List of Tables ix

List of Figures xi

Course Description xiii

1 Data 1
1.1 Baby Crawling Data . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 World Bank Data . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Email Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Handwritten Digit Recognition . . . . . . . . . . . . . . . . . 5
1.5 Looking Forward . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 How to learn (The most important section in this book!) . . 8

2 Introduction to R and RStudio 11

2.1 Obtaining and Installing R . . . . . . . . . . . . . . . . . . . . 11
2.2 Obtaining and Installing RStudio . . . . . . . . . . . . . . . 12
2.3 Using R and RStudio . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Rstudio Themes . . . . . . . . . . . . . . . . . . . . . 13
2.3.2 R as a Calculator . . . . . . . . . . . . . . . . . . . . . 14
2.3.3 Basic descriptive statistics and graphics in R . . . . . 16
2.3.4 An Initial Tour of RStudio . . . . . . . . . . . . . . . 19
2.3.5 Practice Problem . . . . . . . . . . . . . . . . . . . . . 19
2.4 Getting Help . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5 Workspace, Working Directory, and Keeping Organized . . . 20
2.6 Quality of R code . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.6.1 Naming Files . . . . . . . . . . . . . . . . . . . . . . . 22
2.6.2 Naming Variables . . . . . . . . . . . . . . . . . . . . . 22
2.6.3 Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3 Scripts and R Markdown 23

3.1 Scripts in R . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 R Markdown . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.1 Creating and processing R Markdown documents . . . . 31
3.2.2 Text: Lists and Headers . . . . . . . . . . . . . . . . . 32
3.2.3 Code Chunks . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.4 Output formats other than HTML . . . . . . . . . . . 35

iii
iv Contents

3.2.5 LaTeX, knitr, and bookdown . . . . . . . . . . . . . . . 37

3.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4 Data Structures 39
4.1 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.1.1 Types, Conversion, Coercion . . . . . . . . . . . . . . . 41
4.1.2 Accessing Specific Elements of Vectors . . . . . . . . . 44
4.1.3 Practice Problem . . . . . . . . . . . . . . . . . . . . . 46
4.2 Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3 Names of Objects in R . . . . . . . . . . . . . . . . . . . . . 48
4.4 Missing Data, Infinity, etc. . . . . . . . . . . . . . . . . . . . 49
4.4.1 Practice Problem . . . . . . . . . . . . . . . . . . . . . . 51
4.4.2 Infinity and NaN . . . . . . . . . . . . . . . . . . . . . . 51
4.5 Data Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.5.1 Accessing Specific Elements of Data Frames . . . . . . 54
4.6 Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.6.1 Accessing Specific Elements of Lists . . . . . . . . . . 59
4.7 Subsetting with Logical Vectors . . . . . . . . . . . . . . . . . 61
4.7.1 Modifying or Creating Objects via Subsetting . . . . . 62
4.7.2 Logical Subsetting and Data Frames . . . . . . . . . . 63
4.8 Patterned Data . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.8.1 Practice Problem . . . . . . . . . . . . . . . . . . . . . 69
4.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

5 Graphics in R Part 1: ggplot2 71

5.1 Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.1.1 Structure of a Typical ggplot . . . . . . . . . . . . . . 75
5.1.2 Adding lines to a scatter plot . . . . . . . . . . . . . . . 77
5.2 Labels, Axes, Text, etc. . . . . . . . . . . . . . . . . . . . . . 83
5.2.1 Labels . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.3 Customizing Axes . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3.1 Text, Point Size, and Color . . . . . . . . . . . . . . . 86
5.3.2 Practice Problem . . . . . . . . . . . . . . . . . . . . . 88
5.4 Other Types of Graphics . . . . . . . . . . . . . . . . . . . . 90
5.4.1 Histograms . . . . . . . . . . . . . . . . . . . . . . . . 90
5.4.2 Boxplots . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.4.3 Bar Graphs . . . . . . . . . . . . . . . . . . . . . . . . 94
5.4.4 Graphs of Functions . . . . . . . . . . . . . . . . . . . . 97
5.5 Themes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.6 Saving Graphics . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.7 More Resources . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.7.1 Practice Problem . . . . . . . . . . . . . . . . . . . . . 100
5.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

6 Working with Data 103

Contents v

6.1 Reading Data into R . . . . . . . . . . . . . . . . . . . . . . 103

6.2 Reading Data with Missing Observations . . . . . . . . . . . . 107
6.3 Summarizing Data Frames . . . . . . . . . . . . . . . . . . . 109
6.3.1 Column (and Row) Summaries . . . . . . . . . . . . . 109
6.3.2 The apply() Function . . . . . . . . . . . . . . . . . . 112
6.3.3 Practice Problems . . . . . . . . . . . . . . . . . . . . 113
6.3.4 Saving Typing Using with() . . . . . . . . . . . . . . 113
6.4 Transforming a Data Frame . . . . . . . . . . . . . . . . . . 114
6.4.1 Adding Variables . . . . . . . . . . . . . . . . . . . . . 114
6.4.2 Removing Variables . . . . . . . . . . . . . . . . . . . 116
6.4.3 Practice Problem . . . . . . . . . . . . . . . . . . . . . . 117
6.4.4 Transforming Variables . . . . . . . . . . . . . . . . . . 117
6.5 Rearranging Variables . . . . . . . . . . . . . . . . . . . . . . 118
6.6 Reshaping Data . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.6.1 tidyr . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.6.2 Practice Problem . . . . . . . . . . . . . . . . . . . . . 124
6.7 Manipulating Data with dplyr . . . . . . . . . . . . . . . . . 125
6.7.1 Improved Data Frames . . . . . . . . . . . . . . . . . . 126
6.7.2 Filtering Data by Row . . . . . . . . . . . . . . . . . . 128
6.7.3 Selecting variables by column . . . . . . . . . . . . . . 130
6.7.4 Practice Problem . . . . . . . . . . . . . . . . . . . . . 132
6.7.5 Pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.7.6 Arranging Data by Row . . . . . . . . . . . . . . . . . 134
6.7.7 Practice Problem . . . . . . . . . . . . . . . . . . . . . 136
6.7.8 Renaming Variables . . . . . . . . . . . . . . . . . . . . 137
6.7.9 Data Summaries and Grouping . . . . . . . . . . . . . . 137
6.7.10 Creating New Variables . . . . . . . . . . . . . . . . . 142
6.7.11 Practice Problem . . . . . . . . . . . . . . . . . . . . . 145
6.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

7 Functions and Programming 147

7.1 R Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.1.1 Practice Problem . . . . . . . . . . . . . . . . . . . . . . 151
7.1.2 Creating Functions . . . . . . . . . . . . . . . . . . . . . 151
7.2 Programming: Conditional Statements . . . . . . . . . . . . 152
7.2.1 Comparison and Logical Operators . . . . . . . . . . . 153
7.2.2 If else statements . . . . . . . . . . . . . . . . . . . . . 155
7.3 Computer Arithmetic . . . . . . . . . . . . . . . . . . . . . . . 157
7.4 Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.4.1 A Repeat Loop . . . . . . . . . . . . . . . . . . . . . . 159
7.4.2 A While Loop . . . . . . . . . . . . . . . . . . . . . . . 160
7.4.3 A For Loop . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.4.4 Practice Problem . . . . . . . . . . . . . . . . . . . . . . 161
7.5 Efficiency Considerations . . . . . . . . . . . . . . . . . . . . 162
7.5.1 Growing Objects . . . . . . . . . . . . . . . . . . . . . 162
vi Contents

7.5.2 Vectorization . . . . . . . . . . . . . . . . . . . . . . . 164

7.6 More on Functions . . . . . . . . . . . . . . . . . . . . . . . . 166
7.7 Calling Functions . . . . . . . . . . . . . . . . . . . . . . . . 166
7.7.1 The ... argument . . . . . . . . . . . . . . . . . . . . . 167
7.7.2 Lazy Evaluation . . . . . . . . . . . . . . . . . . . . . 168
7.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

8 Spatial Data Visualization and Analysis 171

8.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.1.1 Some Spatial Data Packages . . . . . . . . . . . . . . 172
8.2 Motivating Data . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.3 Reading Spatial Data into R . . . . . . . . . . . . . . . . . . 173
8.4 Coordinate Reference Systems . . . . . . . . . . . . . . . . . 179
8.5 Illustration using ggmap . . . . . . . . . . . . . . . . . . . . . 180
8.6 Illustration using leaflet . . . . . . . . . . . . . . . . . . . 188
8.7 Subsetting Spatial Data . . . . . . . . . . . . . . . . . . . . . 190
8.7.1 Fetching and Cropping Data using raster . . . . . . . 190
8.7.2 Logical, Index, and Name Subsetting . . . . . . . . . . 192
8.7.3 Spatial Subsetting and Overlay . . . . . . . . . . . . . 193
8.7.4 Spatial Aggregation . . . . . . . . . . . . . . . . . . . . 197
8.8 Where to go from here . . . . . . . . . . . . . . . . . . . . . 199
8.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

9 Shiny: Interactive Web Apps in R 201

9.1 Running a Simple Shiny App . . . . . . . . . . . . . . . . . . 202
9.2 Adding User Input . . . . . . . . . . . . . . . . . . . . . . . . 205
9.3 Adding Output . . . . . . . . . . . . . . . . . . . . . . . . . . 206
9.3.1 Interactive Server Logic . . . . . . . . . . . . . . . . . 206
9.3.2 Interactive User Interface . . . . . . . . . . . . . . . . 208
9.4 More Advanced Shiny App: Michigan Campgrounds . . . . . 209
9.4.1 Michigan Campgrounds UI . . . . . . . . . . . . . . . 210
9.4.2 Michigan Campgrounds Server Logic . . . . . . . . . . 210
9.5 Adding Leaflet to Shiny . . . . . . . . . . . . . . . . . . . . . . 211
9.6 Why use Shiny? . . . . . . . . . . . . . . . . . . . . . . . . . 213
9.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

10 Classification 215
10.1 Logistic Regression . . . . . . . . . . . . . . . . . . . . . . . 215
10.1.1 Adding Predictors . . . . . . . . . . . . . . . . . . . . . 221
10.1.2 More than Two Classes . . . . . . . . . . . . . . . . . 224
10.2 Nearest Neighbor Methods . . . . . . . . . . . . . . . . . . . 229
10.2.1 kNN and the Diabetes Data . . . . . . . . . . . . . . . 235
10.2.2 Practice Problem . . . . . . . . . . . . . . . . . . . . . 236
10.2.3 kNN and the iris Data . . . . . . . . . . . . . . . . . . 236
10.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Contents vii

11 Text Data 239

11.1 Reading Text Data into R . . . . . . . . . . . . . . . . . . . 239
11.2 The paste Function . . . . . . . . . . . . . . . . . . . . . . . 244
11.3 More String Processing Functions . . . . . . . . . . . . . . . 249
11.3.1 tolower and toupper . . . . . . . . . . . . . . . . . . 249
11.3.2 nchar and strsplit . . . . . . . . . . . . . . . . . . . 250
11.3.3 Practice Problem . . . . . . . . . . . . . . . . . . . . . 254
11.3.4 nchar Again . . . . . . . . . . . . . . . . . . . . . . . 254
11.3.5 Practice Problem . . . . . . . . . . . . . . . . . . . . . 256
11.3.6 substr and strtrim . . . . . . . . . . . . . . . . . . . . 257
11.3.7 grep and Related Functions . . . . . . . . . . . . . . . 258
11.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

12 Rcpp 261
12.1 Getting Started with Rcpp . . . . . . . . . . . . . . . . . . . . 261
12.1.1 Installation . . . . . . . . . . . . . . . . . . . . . . . . . 261
12.1.2 The Simplest C++ Example . . . . . . . . . . . . . . 262
12.2 Using Rcpp . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
12.2.1 Exporting C++ Functions . . . . . . . . . . . . . . . . 262
12.2.2 Inline C++ Code . . . . . . . . . . . . . . . . . . . . . 264
12.3 The Rcpp Interface . . . . . . . . . . . . . . . . . . . . . . . 265
12.4 No input and scalar output . . . . . . . . . . . . . . . . . . . 265
12.4.1 Scalar input and scalar output . . . . . . . . . . . . . 266
12.4.2 Vector input and scalar output . . . . . . . . . . . . . . 267
12.4.3 Vector input and vector output . . . . . . . . . . . . . 268
12.4.4 Matrix input and vector output . . . . . . . . . . . . . 269
12.4.5 Matrix input and matrix output . . . . . . . . . . . . 270
12.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270

13 Databases and R 271

13.1 SQL and Database Structure . . . . . . . . . . . . . . . . . . . 271
13.1.1 SQL . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
13.2 Difficulties of Working with Large Datasets . . . . . . . . . . 274
13.3 Using dplyr to Query the Database . . . . . . . . . . . . . . 274
13.3.1 Example with RSQLite . . . . . . . . . . . . . . . . . 275
13.3.2 dbplot . . . . . . . . . . . . . . . . . . . . . . . . . . 279
13.3.3 Using Google BigQuery . . . . . . . . . . . . . . . . . . 281
13.4 Changing Records in the Database . . . . . . . . . . . . . . . 286
13.5 R Studio Connections Pane . . . . . . . . . . . . . . . . . . . . 287
13.6 Further Resources . . . . . . . . . . . . . . . . . . . . . . . . . 287
13.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

14 Digital Signal Processing 289

14.1 Introduction to Digital Signal Processing . . . . . . . . . . . 289
14.1.1 Sinusoidal Signals . . . . . . . . . . . . . . . . . . . . 292
viii Contents

14.2 Noise and Filters . . . . . . . . . . . . . . . . . . . . . . . . . 294

14.3 Fourier Transforms and Spectrograms . . . . . . . . . . . . . 298
14.4 Analyzing Audio Data with warbleR . . . . . . . . . . . . . 302
14.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312

15 Graphics in R Part 2: graphics 313

15.1 Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
15.1.1 Adding lines to a scatter plot . . . . . . . . . . . . . . . 317
15.2 Labels, Axes, Text, etc. . . . . . . . . . . . . . . . . . . . . . 322
15.2.1 Labels . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
15.3 Customizing Axes . . . . . . . . . . . . . . . . . . . . . . . . 325
15.3.1 Text, Point Size, and Color . . . . . . . . . . . . . . . . 327
15.4 Other Types of Graphics . . . . . . . . . . . . . . . . . . . . 329
15.4.1 Histograms . . . . . . . . . . . . . . . . . . . . . . . . 329
15.4.2 Boxplots . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
15.4.3 Bar Graphs . . . . . . . . . . . . . . . . . . . . . . . . 333
15.4.4 Graphs of Functions . . . . . . . . . . . . . . . . . . . . 337
15.5 Saving Graphics . . . . . . . . . . . . . . . . . . . . . . . . . 338
15.6 A Summary of Useful graphics Functions and Arguments . 339
15.6.1 Practice Problem . . . . . . . . . . . . . . . . . . . . . 340

16 Solutions to Practice Problems 341

List of Tables

1.1 Data on age at crawling . . . . . . . . . . . . . . . . . . . . . 2

1.2 A small portion of the World Bank data set . . . . . . . . . . 3

4.1 Dimension and type content of base data structures in R . . . 39

8.1 An abbreviated list of ‘sp‘ and ‘raster‘ data objects and associ-
ated classes for the fundamental spatial data types . . . . . . 174

9.1 Interactive elements . . . . . . . . . . . . . . . . . . . . . . . . 211

ix
List of Figures

1.1 A digitized version of a handwritten 6 . . . . . . . . . . . . . 6

1.2 The first 25 handwritten numerals, digitized . . . . . . . . . . . 7
1.3 The first 25 numeral sevens, digitized . . . . . . . . . . . . . . 8

2.1 The RStudio IDE . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2 The RStudio IDE using the Modern theme and the Tomorrow
Night Bright editor theme . . . . . . . . . . . . . . . . . . . . 14
2.3 xkcd: Code Quality . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.1 A script window in RStudio . . . . . . . . . . . . . . . . . . . 29

3.2 Example R Markdown Input and Output . . . . . . . . . . . . 31
3.3 Producing Lists in R Markdown . . . . . . . . . . . . . . . . 32
3.4 Headers and Some LaTeX in R Markdown . . . . . . . . . . . 34
3.5 Other useful LaTeX symbols and expressions in R Markdown 35
3.6 Output of Example R Markdown . . . . . . . . . . . . . . . . 36

8.1 Raster/Vector Comparison [@imageRaster] . . . . . . . . . . 172

9.1 Hello Shiny example app . . . . . . . . . . . . . . . . . . . . . 202

9.2 Running a Shiny app locally . . . . . . . . . . . . . . . . . . . 204
9.3 Your first Shiny app . . . . . . . . . . . . . . . . . . . . . . . 204

14.1 Example of an analog signal: a simple sine curve . . . . . . . 290

14.2 Example of a discrete signal: a sine curve with some missing
values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
14.3 Different available filters within the seewave package . . . . . 299
14.4 Two cedar waxwings, possibly pondering whether or not Ross
and Rachel will ever get back together (oh wait no that’s me as
I write this chapter between binge watching Friends) . . . . . 303
14.5 Example spectrogram of Cedar Waxwing calls . . . . . . . . . . 307
14.6 Example output of autodetec function . . . . . . . . . . . . . 309

xi
Course Description

This book serves as an introduction to programming in R and the use of

associated open source tools. We address practical issues in documenting
workflow, data management, and scientific computing.

xiii
1
Data

Data science is a field that intersects with statistics, mathematics, computer

science, and a wide range of applied fields, such as marketing, biology, and
physics. As such, it is hard to formally define data science, but obviously data
are central to data science, and it is useful at the start to consider some types
of data that are of interest.

1.1 Baby Crawling Data

When thinking about data, we might initially have in mind a modest-sized
and uncomplicated data set consisting primarily of numbers. As an example of
such a data set, a study was done to assess the possible relationship between
the age at which babies first begin to crawl and the temperature at the time
of first crawling. Participants in the study were volunteers.1 The data set
from this study separates the babies by birth month, and reports the birth
month, the average age (in weeks) when first crawling for that month, the
standard deviation of the crawling ages for that month, the number of infants
for that month, and the average temperature during the month when crawling
commenced. The data are shown in Table 1.1 below.2

u <- "https://www.finley-lab.com/files/data/BabyCrawling.tsv"
BabyCrawling <- read.table(u, header=T)

This data set has many simple properties: it is relatively small, there are no
missing observations, the variables are easily understood, etc.
1 More correctly, were volunteered by their parents.
2 These data were retrieved from http://lib.stat.cmu.edu/DASL/Datafiles/Crawling
.html.

1
2 1 Data

TABLE 1.1: Data on age at crawling

BirthMonth AvgCrawlingAge SD n temperature

January 29.84 7.08 32 66
February 30.52 6.96 36 73
March 29.70 8.33 23 72
April 31.84 6.21 26 63
May 28.58 8.07 27 52
June 31.44 8.10 29 39
July 33.64 6.91 21 33
August 32.82 7.61 45 30
September 33.83 6.93 38 33
October 33.35 7.29 44 37
November 33.38 7.42 49 48
December 32.32 5.71 44 57

1.2 World Bank Data

The World Bank provides data related to the development of countries. A data
set was constructed from the World Bank repository. The data set contains
data on countries throughout the world for the years 1960 through 2014 and
contains, among others, variables representing average life expectancy, fertility
rate, and population. Table 1.2 contains the first five records and then 10 more
randomly selected records for these variables in the data set.
Notice that many observations contain missing data for fertility rate and life
expectancy. If all the variables were shown, we would see much more missing
data. This data set is also substantially larger than the baby crawling age data,
with 11880 rows and 15 columns of data in the full data set. (Each column
represents one of the variables. Each row represents one country during one
year).
1.3 Email Data 3

TABLE 1.2: A small portion of the World Bank data set

country year fertility.rate life.expectancy population

Andorra 1978 33746
Andorra 1979 34819
Andorra 1977 32769
Andorra 2007 1.180 81292
Andorra 1976 31781
Cayman Islands 1977 13840
Jamaica 1992 2.855 70.45 2423044
Maldives 1975 6.981 47.97 134077
St. Vincent and the Grenadines 1972 5.629 65.47 92465
Mauritania 1981 6.368 54.84 1578670
Oman 1993 6.098 68.95 2043912
Bahamas, The 1966 3.893 64.68 146364
Angola 1983 7.205 40.53 8489864
French Polynesia 1980 3.989 64.71 151702
Lesotho 1974 5.780 50.29 1122484

1.3 Email Data

It is estimated that in 2015, 90% of the total 205 billion emails sent were spam.3
Spam filters use large amounts of data from emails to learn what distinguishes
spam messages from non-spam (sometimes called “ham”) messages. Below we
include one spam message followed by a ham message.4
From safety33o@l11.newnamedns.com Fri Aug 23 11:03:37 2002
Return-Path: <safety33o@l11.newnamedns.com>
Delivered-To: zzzz@localhost.example.com
Received: from localhost (localhost [127.0.0.1])
by phobos.labs.example.com (Postfix) with ESMTP id 5AC994415F
for <zzzz@localhost>; Fri, 23 Aug 2002 06:02:59 -0400 (EDT)
Received: from mail.webnote.net [193.120.211.219]
by localhost with POP3 (fetchmail-5.9.0)
for zzzz@localhost (single-drop); Fri, 23 Aug 2002 11:02:59 +0100 (IST)
Received: from l11.newnamedns.com ([64.25.38.81])
by webnote.net (8.9.3/8.9.3) with ESMTP id KAA09379
3 RadicatiGroup http://www.radicati.com
4 These messages both come from the large collection of spam and ham messages at
http://spamassassin.apache.org.
4 1 Data

for <zzzz@example.com>; Fri, 23 Aug 2002 10:18:03 +0100

From: safety33o@l11.newnamedns.com
Date: Fri, 23 Aug 2002 02:16:25 -0400
Message-Id: <200208230616.g7N6GOR28438@l11.newnamedns.com>
To: kxzzzzgxlrah@l11.newnamedns.com
Reply-To: safety33o@l11.newnamedns.com
Subject: ADV: Lowest life insurance rates available!
moode

Lowest rates available for term life insurance! Take a moment

and fill out our online form
to see the low rate you qualify for.
Save up to 70% from regular rates! Smokers accepted!
http://www.newnamedns.com/termlife/

Representing quality nationwide carriers. Act now!

From rssfeeds@jmason.org Tue Oct 1 10:37:22 2002
Return-Path: <rssfeeds@example.com>
Delivered-To: yyyy@localhost.example.com
Received: from localhost (jalapeno [127.0.0.1])
by jmason.org (Postfix) with ESMTP id B277816F16
for <jm@localhost>; Tue, 1 Oct 2002 10:37:21 +0100 (IST)
Received: from jalapeno [127.0.0.1]
by localhost with IMAP (fetchmail-5.9.0)
for jm@localhost (single-drop); Tue, 01 Oct 2002 10:37:21 +0100 (IST)
Received: from dogma.slashnull.org (localhost [127.0.0.1]) by
dogma.slashnull.org (8.11.6/8.11.6) with ESMTP id g9180YK15357 for
<jm@jmason.org>; Tue, 1 Oct 2002 09:00:34 +0100
Message-Id: <200210010800.g9180YK15357@dogma.slashnull.org>
To: yyyy@example.com
From: boingboing <rssfeeds@example.com>
Subject: Disney's no-good Park-Czar replaced
Date: Tue, 01 Oct 2002 08:00:34 -0000
Content-Type: text/plain; encoding=utf-8
X-Spam-Status: No, hits=-641.2 required=5.0
tests=AWL
version=2.50-cvs
X-Spam-Level:

URL: http://boingboing.net/#85506723
Date: Not supplied

Disney has named a new president of Walt Disney Parks, replacing Paul Pressler,
the exec who did his damnedest to ruin Disneyland, slashing spending (at the
1.4 Handwritten Digit Recognition 5

expense of safety and employee satisfaction), building the craptastical

California Adventure, reducing the number of SKUs available for sale in the
Park stores, and so on. The new president, James Rasulo, used to be head of
Euro Disney. Link[1] Discuss[2]

[1] http://reuters.com/news_article.jhtml?type=search&StoryID=1510778
[2] http://www.quicktopic.com/boing/H/rw7cDXT3W44C
To implement a spam filter we would have to get the data from these email
messages (and thousands of others) into a software package, extract and
separate potentially important features such as the To: line, the Subject:
line, the message body, etc., and then compare spam and non-spam messages
to find a method to classify new emails correctly. These steps are not simple
in this example. In particular, we would need to become skilled at working
with text data.

1.4 Handwritten Digit Recognition

Correct recognition of handwritten digits by a machine is commonly required
in today’s world. For example, the postal service must scan and recognize zip
codes on handwritten mail. Roughly speaking, a handwritten digit is scanned
and converted to a digital image. To keep things simple we will assume the
scanning creates a grayscale rather than a color image. When converting an
image to a grayscale digital image, a grid of “pixels” is used to represent
the handwritten image, where each pixel has a black intensity value. For
concreteness we’ll assume that intensities are recorded on a scale from −1 (no
black intensity at all) to 1 (maximum black intensity). If the pixel grid is 16
by 16 then the resulting digitized image will contain 256 intensity values, one
for each of the 16 × 16 = 256 pixels.
For example, here are the data corresponding to one handwritten digit, which
happens to be the numeral “6”. Figure 1.1 shows how that digit looks when
digitized.
-1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -0.631 0.862 -0.167
-1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000
-1.000 -1.000 -0.992 0.297 1.000 0.307 -1.000 -1.000 -1.000 -1.000
-1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -0.410 1.000
0.986 -0.565 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000
-1.000 -1.000 -1.000 -0.683 0.825 1.000 0.562 -1.000 -1.000 -1.000
-1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -0.938 0.540
1.000 0.778 -0.715 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000
6 1 Data

-1.000 -1.000 -1.000 -1.000 0.100 1.000 0.922 -0.439 -1.000 -1.000
-1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -0.257
0.950 1.000 -0.162 -1.000 -1.000 -1.000 -0.987 -0.714 -0.832 -1.000
-1.000 -1.000 -1.000 -1.000 -0.797 0.909 1.000 0.300 -0.961 -1.000
-1.000 -0.550 0.485 0.996 0.867 0.092 -1.000 -1.000 -1.000 -1.000
0.278 1.000 0.877 -0.824 -1.000 -0.905 0.145 0.977 1.000 1.000
1.000 0.990 -0.745 -1.000 -1.000 -0.950 0.847 1.000 0.327 -1.000
-1.000 0.355 1.000 0.655 -0.109 -0.185 1.000 0.988 -0.723 -1.000
-1.000 -0.630 1.000 1.000 0.068 -0.925 0.113 0.960 0.308 -0.884
-1.000 -0.075 1.000 0.641 -0.995 -1.000 -1.000 -0.677 1.000 1.000
0.753 0.341 1.000 0.707 -0.942 -1.000 -1.000 0.545 1.000 0.027
-1.000 -1.000 -1.000 -0.903 0.792 1.000 1.000 1.000 1.000 0.536
0.184 0.812 0.837 0.978 0.864 -0.630 -1.000 -1.000 -1.000 -1.000
-0.452 0.828 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
0.135 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -0.483 0.813 1.000
1.000 1.000 1.000 1.000 1.000 0.219 -0.943 -1.000 -1.000 -1.000
-1.000 -1.000 -1.000 -1.000 -0.974 -0.429 0.304 0.823 1.000 0.482
-0.474 -0.991 -1.000 -1.000 -1.000 -1.000

FIGURE 1.1: A digitized version of a handwritten 6

Looking at the digitized images, it may seem simple to correctly identify a

handwritten numeral. But remember, the machine only has access to the 256
pixel intensities, and must make a decision based on them.
Figure 1.2 shows the digitized images of the first 25 numerals in the data set,
1.5 Looking Forward 7

and Figure 1.3 shows the digitized images of the first 25 numeral sevens in the
data set. These give some idea of the variability in how digits are written.5

FIGURE 1.2: The first 25 handwritten numerals, digitized

1.5 Looking Forward

The four examples above illustrate a small sample of the wide variety of data
sets that may be encountered in data science. Each of these provides its own
challenges. The baby crawling data present challenges that are more statistical
in nature. For example, how might the study design (which isn’t described
here) affect methods of analysis and conclusions drawn from the study? Similar
challenges are also present within the other data sets, but these data sets also
present more substantial challenges prior to (and during) the analysis stage,
such as how to work with the missing data in the World Bank data set, or
how to effectively and efficiently process the email data to extract features of
interest.
This book and associated material introduce tools to tackle some of the
5 Actually, these data were already pre-processed to get the orientation correct. Actual

handwritten digits would be even more variable.

8 1 Data

FIGURE 1.3: The first 25 numeral sevens, digitized

challenges in working with real data sets, within the context of the R statistical
system. We will focus on important topics such as

1. Obtaining and manipulating data

2. Graphical tools for exploring and summarizing data
3. Communicating findings about data that support reproducible re-
search
4. Tools for classification problems such as email spam filtering or
handwritten digit recognition
5. Programming and writing functions in R

1.6 How to learn (The most important section in this

book!)
There are several ways to engage with the content of this book and associated
materials.
One way is not to engage at all. Leave the book closed on a shelf and do
1.6 How to learn (The most important section in this book!) 9

something else with your time. That may or may not be a good life strategy,
depending on what else you do with your time, but you won’t learn much from
the book!
Another way to engage is to read through the book “passively”, reading all
that’s written but not reading the book while at your computer, where you
could enter the R commands from the book. With this strategy you’ll probably
learn more than if you leave the book closed on a shelf, but there are better
options.
A third way to engage is to read the book while you’re at a computer, enter
the R commands from the book as you read about them, and work on the
practice problems within many of the chapters. You’ll likely learn more this
way.
A fourth strategy is even better. In addition to reading, entering the commands
given in the book, and working through the practice exercises, you think about
what you’re doing, and ask yourself questions (which you then go on to answer).
For example after working through some R code computing the logarithm of
positive numbers you might ask yourself, “What would R do if I asked it to
calculate the logarithm of a negative number? What would R do if I asked it
to calculate the logarithm of a really large number such as one trillion?” You
could explore these questions easily by just trying things out in the R Console
window.
If your goal is to maximize the time you have to binge-watch Stranger Things
Season 2 on Netflix, the first strategy may be optimal. But if your goal is to
learn a lot about computational tools for data science, the fourth strategy is
probably going to be best.
2
Introduction to R and RStudio

Various statistical and programming software environments are used in data

science, including R, Python, SAS, C++, SPSS, and many others. Each has
strengths and weaknesses, and often two or more are used in a single project.
This book focuses on R for several reasons:

1. R is free
2. It is one of, if not the, most widely used software environments in
data science
3. R is under constant and open development by a diverse and expert
core group
4. It has an incredible variety of contributed packages
5. A new user can (relatively) quickly gain enough skills to obtain,
manage, and analyze data in R

Several enhanced interfaces for R have been developed. Generally such interfaces
are referred to as integrated development environments (IDE). These interfaces
are used to facilitate software development. At minimum, an IDE typically
consists of a source code editor and build automation tools. We will use the
RStudio IDE, which according to its developers “is a powerful productive
user interface for R.”1 RStudio is widely used, it is used increasingly in the R
community, and it makes learning to use R a bit simpler. Although we will
use RStudio, most of what is presented in this book can be accomplished in R
(without an added interface) with few or no changes.

2.1 Obtaining and Installing R

It is simple to install R on computers running Microsoft Windows, macOS, or
Linux. For other operating systems users can compile the source code directly.2
1 http://www.rstudio.com/
2 Windows, macOS, and Linux users also can compile the source code directly, but for

most it is a better idea to install R from already compiled binary distributions.

11
12 2 Introduction to R and RStudio

Here is a step-by-step guide to installing R for Microsoft Windows.3 macOS

and Linux users would follow similar steps.

1. Go to http://www.r-project.org/
2. Click on the CRAN link on the left side of the page
3. Choose one of the mirrors.4
4. Click on Download R for Windows
5. Click on base
6. Click on Download R 4.0.2 for Windows
7. Install R as you would install any other Windows program

2.2 Obtaining and Installing RStudio

You must install R prior to installing RStudio. RStudio is also simple to install:

1. Go to http://www.rstudio.com
2. Click on the link RStudio under the Products tab
3. Click on the RStudio Desktop box
4. Choose the DOWNLOAD RSTUDIO DESKTOP link in the Open Source
Edition column
5. On the ensuing page, click on the Installer version for your oper-
ating system, and once downloaded, install as you would any other
program

2.3 Using R and RStudio

Start RStudio as you would any other program in your operating system. For
example, under Microsoft Windows use the Start Menu or double click on the
shortcut on the desktop (if a shortcut was created in the installation process).
A (rather small) view of RStudio is displayed in Figure 2.1.
Initially the RStudio window contains three smaller windows. For now our
main focus will be the large window on the left, the Console window, in which
3 New versions of R are released regularly, so the version number in Step 6 might be

different from what is listed below.

4 The https://cloud.r-project.org/ mirror is usually fast. Otherwise choose a mirror

in Michigan.
2.3 Using R and RStudio 13

FIGURE 2.1: The RStudio IDE

R statements are typed. The next few sections give simple examples of the use
of R. In these sections we will focus on small and non-complex data sets, but
of course later in the book we will work with much larger and more complex
sets of data. Read these sections at your computer with R running, and enter
the R commands there to get comfortable using the R console window and
RStudio.

2.3.1 Rstudio Themes

Figure 2.1 shows the default Rstudio theme and layout. RStudio provides great
flexibility in changing themes and code highlighting to customize the RStudio
interface. To switch between themes, click on the Tools bar at the top of the
window, click on the Global Options tab, and then click on the Appearance
tab where you will see the option to switch between three RStudio themes:
Modern, Classic, and Sky.
You can also switch the Editor font as well as specify the Editor theme from
a suite of options. Some of the editor themes are dark palettes that will
automatically activate a Dark theme. For example, an RStudio window using
the Modern theme and the Tomorrow Night Bright editor theme is shown
in Figure 2.2. Using a dark palette in RStudio can be especially nice if you
spend numerous hours at a time looking at your screen. If you’re interested in
14 2 Introduction to R and RStudio

specializing your Rstudio window even further, you can even create custom
editor themes (see this Rstudio blog5 for a useful tutorial).

FIGURE 2.2: The RStudio IDE using the Modern theme and the Tomorrow
Night Bright editor theme

2.3.2 R as a Calculator

R can be used as a calculator. Note that # is the comment character in R, so

R ignores everything following this character. Also, you will see that R prints
[1] before the results of each command. Soon we will explain its relevance,
but ignore this for now. The command prompt in R is the greater than sign >.

34+20*sqrt(100) ## +,-,*,/ have the expected meanings

[1] 234

exp(2) ##The exponential function

[1] 7.389

log10(100) ##Base 10 logarithm

5 https://support.rstudio.com/hc/en-us/articles/115011846747-Using-RStudio-

Themes
2.3 Using R and RStudio 15

[1] 2

log(100) ##Base e logarithm

[1] 4.605

10ˆlog10(55)

[1] 55
Most functions in R can be applied to vector arguments rather than operating
on a single argument at a time. A vector is a data structure that contains
elements of the same data type (i.e., integers).

1:25 ##The integers from 1 to 25

[1] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
[18] 18 19 20 21 22 23 24 25

log(1:25) ##The base e logarithm of these integers

[1] 0.0000 0.6931 1.0986 1.3863 1.6094 1.7918 1.9459

[8] 2.0794 2.1972 2.3026 2.3979 2.4849 2.5649 2.6391
[15] 2.7081 2.7726 2.8332 2.8904 2.9444 2.9957 3.0445
[22] 3.0910 3.1355 3.1781 3.2189

1:25*1:25 ##What will this produce?

[1] 1 4 9 16 25 36 49 64 81 100 121 144

[13] 169 196 225 256 289 324 361 400 441 484 529 576
[25] 625

1:25*1:5 ##What about this?

[1] 1 4 9 16 25 6 14 24 36 50 11 24
[13] 39 56 75 16 34 54 76 100 21 44 69 96
[25] 125

seq(from=0, to=1, by=0.1) ##A sequence of numbers from 0 to 1

[1] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
16 2 Introduction to R and RStudio

exp(seq(from=0, to=1, by=0.1)) ##What will this produce?

[1] 1.000 1.105 1.221 1.350 1.492 1.649 1.822 2.014

[9] 2.226 2.460 2.718
Now the mysterious square bracketed numbers appearing next to the output
make sense. R puts the position of the beginning value on a line in square
brackets before the line of output. For example if the output has 40 values,
and 15 values appear on each line, then the first line will have [1] at the left,
the second line will have [16] to the left, and the third line will have [31] to
the left.

2.3.3 Basic descriptive statistics and graphics in R

It is easy to compute basic descriptive statistics and to produce standard

graphical representations of data in R. First we create three variables with
horsepower, miles per gallon, and names for 15 cars.6 In this case with a
small data set we enter the data “by hand” using the c() function, which
concatenates its arguments into a vector. For larger data sets we will clearly
want an alternative. Note that character values are surrounded by quotation
marks.
A style note: R has two widely used methods of assignment: the left arrow,
which consists of a less than sign followed immediately by a dash: <- and the
equals sign: =. Much ink has been used debating the relative merits of the two
methods, and their subtle differences. Many leading R style guides (e.g., the
Google style guide at https://google.github.io/styleguide/Rguide.xml
and the Bioconductor style guide at http://www.bioconductor.org/develop
ers/how-to/coding-style/) recommend the left arrow <- as an assignment
operator, and we will use this throughout the book.
Also you will see that if a command has not been completed but the ENTER
key is pressed, the command prompt changes to a + sign. To get back to the
regular prompt sign, you can either type something to finish the command
(i.e., ) or ]), or you can press your ESC button and retype the whole command.

car.hp <- c(110, 110, 93, 110, 175, 105, 245, 62, 95, 123,
123, 180, 180, 180, 205)
car.mpg <- c(21.0, 21.0, 22.8, 21.4, 18.7, 18.1, 14.3, 24.4,
22.8, 19.2, 17.8, 16.4, 17.3, 15.2, 10.4)
car.name <- c("Mazda RX4", "Mazda RX4 Wag", "Datsun 710",
"Hornet 4 Drive", "Hornet Sportabout", "Valiant",

6 These are from a relatively old data set, with 1974 model cars.
2.3 Using R and RStudio 17

"Duster 360", "Merc 240D", "Merc 230", "Merc 280",

"Merc 280C", "Merc 450SE", "Merc 450SL",
"Merc 450SLC", "Cadillac Fleetwood")
car.hp

[1] 110 110 93 110 175 105 245 62 95 123 123 180
[13] 180 180 205

car.mpg

[1] 21.0 21.0 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2
[11] 17.8 16.4 17.3 15.2 10.4

car.name

[1] "Mazda RX4" "Mazda RX4 Wag"

[3] "Datsun 710" "Hornet 4 Drive"
[5] "Hornet Sportabout" "Valiant"
[7] "Duster 360" "Merc 240D"
[9] "Merc 230" "Merc 280"
[11] "Merc 280C" "Merc 450SE"
[13] "Merc 450SL" "Merc 450SLC"
[15] "Cadillac Fleetwood"
Next we compute some descriptive statistics for the two numeric variables
(car.hp and car.mpg)

mean(car.hp)

[1] 139.7

sd(car.hp)

[1] 50.78

summary(car.hp)

Min. 1st Qu. Median Mean 3rd Qu. Max.

62 108 123 140 180 245

mean(car.mpg)
18 2 Introduction to R and RStudio

[1] 18.72

sd(car.mpg)

[1] 3.714

summary(car.mpg)

Min. 1st Qu. Median Mean 3rd Qu. Max.

10.4 16.9 18.7 18.7 21.2 24.4
Next, a scatter plot of cars.mpg versus cars.hp:

plot(car.hp, car.mpg)
18 20 22 24
car.mpg

10 12 14 16

100 150 200 250

car.hp

Unsurprisingly as horsepower increases, mpg tends to decrease. This relation-

ship can be investigated further using linear regression, a statistical procedure
that involves fitting a linear model to a data set in order to further understand
the relationship between two variables.
2.4 Getting Help 19

2.3.4 An Initial Tour of RStudio

When you created the car.hp and other vectors in the previous section, you
might have noticed the vector name and a short description of its attributes
appear in the top right Global Environment window. Similarly, when you
called plot(car.hp,car.mpg) the corresponding plot appeared in the lower
right Plots window.
A comprehensive, but slightly overwhelming, cheatsheet for RStudio is avail-
able here https://www.rstudio.com/wp-content/uploads/2016/01/rstu
dio-IDE-cheatsheet.pdf. As we progress in learning R and RStudio, this
cheatsheet will become more useful. For now you might use the cheatsheet to
locate the various windows and functions identified in the coming chapters.

2.3.5 Practice Problem

When running a large program consisting of numerous lines of complicated

code, it is often basic algebraic typos that lead to the most frustrating bugs.
Having a good grip on the order of operations is a basic, yet very important
skill for writing good code. To practice, compute the following operation in R.
√
(e14 + log10 (8)) × 5
loge (4) − 5 ∗ 102

2.4 Getting Help

There are several free (and several not free) ways to get R help when needed.
Several help-related functions are built into R. If there’s a particular R function
of interest, such as log, help(log) or ?log will bring up a help page for that
function. In RStudio the help page is displayed, by default, in the Help tab
in the lower right window.7 The function help.start opens a window which
allows browsing of the online documentation included with R. To use this,
type help.start() in the console window.8 The help.start function also
provides several manuals online and can be a useful interface in addition to
the built in help.
7 There are ways to change this default behavior.
8 You may wonder about the parentheses after help.start. A user can specify arguments
to any R function inside parentheses. For example log(10) asks R to return the logarithm
of the argument 10. Even if no arguments are needed, R requires empty parentheses at the
end of any function name. In fact if you just type the function name without parentheses, R
returns the definition of the function. For simple functions this can be illuminating.
20 2 Introduction to R and RStudio

Search engines provide another, sometimes more user-friendly, way to receive

answers for R questions. A Google search often quickly finds something written
by another user who had the same (or a similar) question, or an online tutorial
that touches on the question. More specialized is rseek.org9 , which is a search
engine focused specifically on R. Both Google and rseek.org are valuable tools,
often providing more user-friendly information than R’s own help system.
In addition, R users have written many types of contributed documentation.
Some of this documentation is available at http://cran.r-project.org/ot
her-docs.html. Of course there are also numerous books covering general
and specialized R topics available for purchase.

2.5 Workspace, Working Directory, and Keeping Orga-

nized
The workspace is your R session working environment and includes any objects
you create. Recall these objects are listed in the Global Environment window.
The command ls(), which stands for list, will also list all the objects in your
workspace (note, this is the same list that is given in the Global Environment
window). When you close RStudio, a dialog box will ask you if you want to
save an image of the current workspace. If you choose to save your workspace,
RStudio saves your session objects and information in a .RData file (the period
makes it a hidden file) in your working directory. Next time you start R or
RStudio it checks if there is a .RData in the working directory, loads it if
it exists, and your session continues where you left off. Otherwise R starts
with an empty workspace. This leads to the next question—what is a working
directory?
Each R session is associated with a working directory. This is just a directory
from which R reads and writes files, e.g., the .RData file, data files you want
to analyze, or files you want to save. On Mac when you start RStudio it sets
the working directory to your home directory (for me that’s /Users/andy). If
you’re on a different operating system, you can check where the default working
directory is by typing getwd() in the console. You can change the default
working directory under RStudio’s Global Option dialog found under the
Tools dropdown menu. There are multiple ways to change the working directory
once an R session is started in RStudio. One method is to click on the Files
tab in the lower right window and then click the More button. Alternatively,
you can set the session’s working directory using the setwd() in the console.
For example, on Windows setwd("C:/Users/andy/for875/exercise1") will
9 http://rseek.org
2.6 Quality of R code 21

set the working directory to C:/Users/andy/for875/exercise1, assuming

that file path and directory exist (Note: Windows file path uses a backslash, \,
but in R the backslash is an escape character, hence specifying file paths in
R on Windows uses the forward slash, i.e., /). Similarly on Mac you can use
setwd("/Users/andy/for875/exercise1"). Perhaps the most simple method
is to click on the Session tab at the top of your screen and click on the Set
Working Directory option. Later on when we start reading and writing data
from our R session, it will be very important that you are able to identify
your current working directory and change it if needed. We will revisit this in
subsequent chapters.
As with all work, keeping organized is the key to efficiency. It is good practice
to have a dedicated directory for each R project or exercise.

2.6 Quality of R code

FIGURE 2.3: xkcd: Code Quality

Writing well-organized and well-labeled code allows your code to be more

easily read and understood by another person. (See xkcd’s take on code quality
in Figure 2.3.) More importantly, though, your well-written code is more
accessible to you hours, days, or even months later. We are hoping that you
can use the code you write in this class in future projects and research.
Google provides style guides for many programming languages. You can find
the R style guide here10 . Below are a few of the key points from the guide that
we will use right away.
10 https://google.github.io/styleguide/Rguide.xml
 22 2 Introduction to R and RStudio

2.6.1 Naming Files

File names should be meaningful and end in .R. If we write a script that
analyzes a certain species distribution:
• GOOD: african_rhino_distribution.R
• GOOD: africanRhinoDistribution.R
• BAD: speciesDist.R (too ambiguous)
• BAD: species.dist.R (too ambiguous and two periods can confuse operat-
ing systems’ file type auto-detect)
• BAD: speciesdist.R (too ambiguous and confusing)

2.6.2 Naming Variables

• GOOD: rhino.count
• GOOD: rhinoCount
• GOOD: rhino_count (We don’t mind the underscore and use it quite often,
although Google’s style guide says it’s a no-no for some reason)
• BAD: rhinocount (confusing)

2.6.3 Syntax

• Keep code lines under 80 characters long.

• Indent your code with two spaces. (RStudio does this by default when you
press the TAB key.)
3
Scripts and R Markdown

Doing work in data science, whether for homework, a project for a business, or
a research project, typically involves several iterations. For example creating
an effective graphical representation of data can involve trying out several
different graphical representations, and then tens if not hundreds of iterations
when fine-tuning the chosen representation. Furthermore, each of these repre-
sentations may require several R commands to create. Although this all could
be accomplished by typing and re-typing commands at the R Console, it is
easier and more effective to write the commands in a script file, which then
can be submitted to the R console either a line at a time or all together.1
In addition to making the workflow more efficient, R scripts provide another
large benefit. Often we work on one part of a homework assignment or project
for a few hours, then move on to something else, and then return to the original
part a few days, months, or sometimes even years later. In such cases we may
have forgotten how we created the graphical display that we were so proud of,
and will need to again spend a few hours to recreate it. If we save a script file,
we have the ingredients immediately available when we return to a portion of
a project.2
Next consider the larger scientific endeavor. Ideally a scientific study will be
reproducible, meaning that an independent group of researchers (or the original
researchers) will be able to duplicate the study. Thinking about data science,
this means that all the steps taken when working with the data from a study
should be reproducible, from the selection of variables to formal data analysis.
In principle, this can be facilitated by explaining, in words, each step of the
work with data. In practice, it is typically difficult or impossible to reproduce
a full data analysis based on a written explanation. Much more effective is
to include the actual computer code which accomplished the data work in
the report, whether the report is a homework assignment or a research paper.
Tools in R such as R Markdown facilitate this process.
1 Unsurprisingly it is also possible to submit several selected lines of code at once.
2 Inprinciple the R history mechanism provides a similar record. But with history we
have to search through a lot of other code to find what we’re looking for, and scripts are a
much cleaner mechanism to record our work.

23
24 3 Scripts and R Markdown

3.1 Scripts in R
As noted above, scripts help to make work with data more efficient and provide
a record of how data were managed and analyzed. Below we describe an
example. This example uses features of R that we have not yet discussed, so
don’t worry about the details, but rather about how it motivates the use of a
script file.
First we read in a data set containing data on (among other things) fertility
rate and life expectancy for countries throughout the world, for the years 1960
through 2014.

u <- "https://www.finley-lab.com/files/data/WorldBank.csv"
WorldBank <- read.csv(u, header=TRUE, stringsAsFactors=FALSE)

Next we print the names of the variables in the data set. Don’t be concerned
about the specific details. Later we will learn much more about reading in
data and working with data sets in R.

names(WorldBank)

[1] "iso2c"
[2] "country"
[3] "year"
[4] "fertility.rate"
[5] "life.expectancy"
[6] "population"
[7] "GDP.per.capita.Current.USD"
[8] "X15.to.25.yr.female.literacy"
[9] "iso3c"
[10] "region"
[11] "capital"
[12] "longitude"
[13] "latitude"
[14] "income"
[15] "lending"
We will try to create a scatter plot of fertility rate versus life expectancy of
countries for the year 1960. To do this we’ll first create variables containing
the values of fertility rate and life expectancy for 19603 , and print out the first
ten values of each variable.

3 This isn’t necessary, but it is convenient

3.1 Scripts in R 25

fertility <- WorldBank$fertility.rate[WorldBank$year == 1960]

lifeexp <- WorldBank$life.expectancy[WorldBank$year==1960]
fertility[1:10]

[1] NA 6.928 7.671 4.425 6.186 4.550 7.316 3.109

[9] NA 2.690

lifeexp[1:10]

[1] NA 52.24 31.58 61.78 62.25 65.86 32.98 65.22

[9] NA 68.59
We see that some countries do not have data for 1960. R represents missing
data via NA. Of course at some point it would be good to investigate which
countries’ data are missing and why. The plot() function in R will just omit
missing values, and for now we will just plot the non-missing data. A scatter
plot of the data is drawn next.

plot(lifeexp, fertility)
8
7
6
fertility

5
4
3
2

30 40 50 60 70

lifeexp

The scatter plot shows that as life expectancy increases, fertility rate tends to
decrease in what appears to be a nonlinear relationship. Now that we have a
basic scatter plot, it is tempting to make it more informative. We will do this
26 3 Scripts and R Markdown

by adding two features. One is to make the points’ size proportional to the
country’s population, and the second is to make the points’ color represent the
region of the world the country resides in. We’ll first extract the population
and region variables for 1960.

pop <- WorldBank$population[WorldBank$year==1960]

region <- WorldBank$region[WorldBank$year==1960]
pop[1:10]

[1] 13414 89608 8774440 54681 1608800

[6] 1867396 4965988 20623998 20012 7047539

region[1:10]

[1] "Europe & Central Asia (all income levels)"

[2] "Middle East & North Africa (all income levels)"
[3] "South Asia"
[4] "Latin America & Caribbean (all income levels)"
[5] "Europe & Central Asia (all income levels)"
[6] "Europe & Central Asia (all income levels)"
[7] "Sub-Saharan Africa (all income levels)"
[8] "Latin America & Caribbean (all income levels)"
[9] "East Asia & Pacific (all income levels)"
[10] "Europe & Central Asia (all income levels)"
To create the scatter plot we will do two things. First we will create the axes,
labels, etc. for the plot, but not plot the points. The argument type="n" tells
R to do this. Then we will use the symbols() function to add symbols, the
circles argument to set the sizes of the points, and the bg argument to set
the colors. Don’t worry about the details! In fact, later in the book we will
learn about an R package called ggplot2 that provides a different way to
create such plots. You’ll see two plots below, first the “empty” plot which is
just a building block, then the plot including the appropriate symbols.

plot(lifeexp, fertility, type="n")

3.1 Scripts in R 27

8
7
6
fertility

5
4
3
2

30 40 50 60 70

lifeexp

symbols(lifeexp, fertility, circles=sqrt(pop/pi), inches=0.35,

bg=match(region, unique(region)))
8
6
fertility

4
2

30 40 50 60 70 80

lifeexp

Of course we should have a key which tells the viewer which region each color
28 3 Scripts and R Markdown

represents, and a way to determine which country each point represents, and
a lot of other refinements. For now we will resist such temptations.
Some of the process leading to the completed plot is shown above, such as
reading in the data, creating variables representing the 1960 fertility rate and
life expectancy, an intermediate plot that was rejected, and so on. A lot of
the process isn’t shown, simply to save space. There would likely be mistakes
(either minor typing mistakes or more complex errors). Focusing only on the
symbols() function that was used to add the colorful symbols to the scatter
plot, there would likely have been a substantial number of attempts with
different values of the circles, inches, and bg arguments before settling on
the actual form used to create the plot. This is the typical process you will
soon discover when producing useful data visualizations.
Now imagine trying to recreate the plot a few days later. Possibly someone saw
the plot and commented that it would be interesting to see some similar plots,
but for years in the 1970s when there were major famines in different countries
of the world. If all the work, including all the false starts and refinements, were
done at the console, it would be hard to sort things out and would take longer
than necessary to create the new plots. This would be especially true if a few
months had passed rather than just a few days.
But with a script file, especially a script file with a few well-chosen comments,
creating the new scatter plots would be much easier. Fortunately it is quite
easy to create and work with script files in RStudio.4 Just choose File > New
File > R script and a script window will open up in the upper left of the
full RStudio window.
An example of a script window (with some R code already typed in) is shown
in Figure 3.1. From the script window the user can, among other things, save
the script (either using the File menu or the icon near the top left of the
window) and run one or more lines of code from the window (using the run
icon in the window, or by copying and pasting into the console window). In
addition, there is a Source on Save checkbox. If this is checked, the R code
in the script window is automatically read into R and executed when the script
file is saved.
4 It is also easy in R without RStudio. Just use File > New script to create a script file,

and save it before exiting R.

3.2 R Markdown 29

FIGURE 3.1: A script window in RStudio

3.2 R Markdown
People typically work on data with a larger purpose in mind. Possibly the
purpose is to understand a biological system more clearly. Possibly the purpose
is to refine a system that recommends movies to users in an online streaming
movie service. Possibly the purpose is to complete a homework assignment
and demonstrate to the instructor an understanding of an aspect of data
analysis. Whatever the purpose, a key aspect is communicating with the
desired audience.
One possibility, which is somewhat effective, is to write a document using
software such as Microsoft Word 5 and to include R output such as computations
and graphics by cutting and pasting into the main document. One drawback
to this approach is similar to what makes script files so useful: If the document
must be revised it may be hard to unearth the R code that created graphics
or analyses.6 A more subtle but possibly more important drawback is that
the reader of the document will not know precisely how analyses were done,
or how graphics were created. Over time even the author(s) of the paper will
forget the details. A verbal description in a “methods” section of a paper can
help here, but typically these do not provide all the details of the analysis,
5 Or possibly LaTeX if the document is more technical
6 Organizing the R code using script files and keeping all the work organized in a well-
thought-out directory structure can help here, but this requires a level of forethought and
organization that most people do not possess . . . including myself.
30 3 Scripts and R Markdown

but rather might state something like, “All analyses were carried out using R
version 3.4.0.”
RStudio’s website provides an excellent overview of R Markdown capabilities
for reproducible research. At minimum, follow the Get Started link at http:
//rmarkdown.rstudio.com/ and watch the introduction video.
Among other things, R Markdown provides a way to include R code that reads
in data, creates graphics, or performs analyses, all in a single document that is
processed to create a research paper, homework assignment, or other written
product. The R Markdown file is a plain text file containing text the author
wants to show in the final document, simple commands to indicate how the
text should be formatted (for example boldface, italic, or a bulleted list), and R
code that creates output (including graphics) on the fly. Perhaps the simplest
way to get started is to see an R Markdown file and the resulting document
that is produced after the R Markdown document is processed. In Figure 3.2
we show the input and output of an example R Markdown document. In this
case the output created is an HTML file, but there are other possible output
formats, such as Microsoft Word or PDF.
At the top of the input R Markdown file are some lines with --- at the top and
the bottom. These lines are not needed, but give a convenient way to specify
the title, author, and date of the article that are then typeset prominently at
the top of the output document. For now, don’t be concerned with the lines
following output:. These can be omitted (or included as shown).
Next are a few lines showing some of the ways that font effects such as italics,
boldface, and strikethrough can be achieved. For example, an asterisk before
and after text sets the text in italics, and two asterisks before and after text
sets the text in boldface.
More important for our purposes is the ability to include R code in the R
Markdown file, which will be executed with the output appearing in the
output document. Bits of R code included this way are called code chunks. The
beginning of a code chunk is indicated with three backticks and an “r” in curly
braces: ```{r}. The end of a code chunk is indicated with three backticks ```.
For example, the R Markdown file in Figure 3.2 has one code chunk:

```{r}
x = 1:10
y = 10:1
mean(x)
sd(y)
```

In this code chunk two vectors x and y are created, and the mean of x and the
standard deviation of y are computed. In the output in Figure 3.2 the R code
3.2 R Markdown 31

FIGURE 3.2: Example R Markdown Input and Output

is reproduced, and the output of the two lines of code asking for the mean and
standard deviation is shown.

3.2.1 Creating and processing R Markdown documents

RStudio has features which facilitate creating and processing R Markdown

documents. Choose File > New File > R Markdown.... In the ensuing di-
alog box, make sure that Document is highlighted on the left, enter the title
and author (if desired), and choose the Default Output Format (HTML is
good to begin). Then click OK. A document will appear in the upper left of
the RStudio window. It is an R Markdown document, and the title and author
you chose will show up, delimited by --- at the top of the document. A generic
body of the document will also be included.
For now just keep this generic document as is. To process it to create the HTML
 32 3 Scripts and R Markdown

output, click the Knit HTML button at the top of the R Markdown window7 .
You’ll be prompted to choose a filename for the R Markdown file. Make sure
that you use .Rmd as the extension for this file. Once you’ve successfully saved
the file, RStudio will process the file, create the HTML output, and open this
output in a new window. The HTML output file will also be saved to your
working directory. This file can be shared with others, who can open it using a
web browser such as Chrome or Firefox.
There are many options which allow customization of R Markdown documents.
Some of these affect formatting of text in the document, while others affect
how R code is evaluated and displayed. The RStudio web site contains a
useful summary of many R Markdown options at https://www.rstudio.
com/wp-content/uploads/2015/03/rmarkdown-reference.pdf. A different,
but mind-numbingly busy, cheatsheet is at https://www.rstudio.com/wp-
content/uploads/2016/03/rmarkdown-cheatsheet-2.0.pdf. Some of the
more commonly used R Markdown options are described next.

3.2.2 Text: Lists and Headers

Unordered (sometimes called bulleted) lists and ordered lists are easy in R
Markdown. Figure 3.3 illustrates the creation of unordered and ordered lists.

FIGURE 3.3: Producing Lists in R Markdown

• For an unordered list, either an asterisk, a plus sign, or a minus sign may
7 If you hover your mouse over this Knit button after a couple seconds it should display a

keyboard shortcut for you to do this if you don’t like pushing buttons
 3.2 R Markdown 33

precede list items. Use a space after these symbols before including the
list text. To have second-level items (sub-lists) indent four spaces before
indicating the list item. This can also be done for third-level items.
• For an ordered list use a numeral followed by a period and a space (1. or
2. or 3. or . . . ) to indicate a numbered list, and use a letter followed by a
period and a space (a. or b. or c. or . . . ) to indicate a lettered list. The same
four space convention used in unordered lists is used to designate ordered
sub lists.
• For an ordered list, the first list item will be labeled with the number or letter
that you specify, but subsequent list items will be numbered sequentially.
The example in Figure 3.3 will make this more clear. In those examples
notice that for the ordered list, although the first-level numbers given in the
R Markdown file are 1, 2, and 17, the numbers printed in the output are 1,
2, and 3. Similarly the letters given in the R Markdown file are c and q, but
the output file prints c and d.
R Markdown does not give substantial control over font size. Different “header”
levels are available that provide different font sizes. Put one or more hash
marks in front of text to specify different header levels. Other font choices such
as subscripts and superscripts are possible, by surrounding the text either by
tildes or carets. More sophisticated mathematical displays are also possible,
and are surrounded by dollar signs. The actual mathematical expressions are
specified using a language called LaTeX See Figures 3.4 and 3.5 for examples.

3.2.3 Code Chunks

R Markdown provides a large number of options to vary the behavior of code

chunks. In some contexts it is useful to display the output but not the R code
leading to the output. In some contexts it is useful to display the R prompt,
while in others it is not. Maybe we want to change the size of figures created
by graphics commands. And so on. A large number of code chunk options
are described in http://www.rstudio.com/wp-content/uploads/2015/03/
rmarkdown-reference.pdf.
Code chunk options are specified in the curly braces near the beginning of a
code chunk. Below are a few of the more commonly used options. The use of
these options is illustrated in Figure 3.6.

1. echo=FALSE specifies that the R code itself should not be printed,

but any output of the R code should be printed in the resulting
document.
2. include=FALSE specifies that neither the R code nor the output
34 3 Scripts and R Markdown

FIGURE 3.4: Headers and Some LaTeX in R Markdown

should be printed. However, the objects created by the code chunk

will be available for use in later code chunks.
3. eval=FALSE specifies that the R code should not be evaluated. The
code will be printed unless, for example, echo=FALSE is also given
as an option.
4. error=FALSE and warning=FALSE specify that, respectively, error
messages and warning messages generated by the R code should not
be printed.
5. The comment option allows a specified character string to be
prepended to each line of results. By default this is set to comment
= '##' which explains the two hash marks preceding the results
in Figure 3.2. Setting comment = NA presents output without any
character string prepended. That is done in most code chunks in
this book.
3.2 R Markdown 35

FIGURE 3.5: Other useful LaTeX symbols and expressions in R Markdown

6. prompt=TRUE specifies that the R prompt > will be prepended to each

line of R code shown in the document. prompt = FALSE specifies
that command prompts should not be included.
7. fig.height and fig.width specify the height and width of figures
generated by R code. These are specified in inches. For example,
fig.height=4 specifies a four inch high figure.

Figures 3.6 gives examples of the use of code chunk options.

3.2.4 Output formats other than HTML

It is possible to use R Markdown to produce documents in formats other than

HTML, including Word and PDF documents. Next to the Knit HTML button
is a down arrow. Click on this and choose Knit Word to produce a Microsoft
word output document. Although there is also a Knit PDF button, PDF output
requires additional software called TeX in addition to RStudio.8
8 It isn’t particularly hard to install TeX software. For a Microsoft Windows system,

MiKTeX is convenient and is available from https://miktex.org. For a Mac system, MacTeX
is available from https://www.tug.org/mactex/
36 3 Scripts and R Markdown

FIGURE 3.6: Output of Example R Markdown

3.3 Exercises 37

3.2.5 LaTeX, knitr, and bookdown

While basic R Markdown provides substantial flexibility and power, it lacks

features such as cross-referencing, fine control over fonts, etc. If this is desired,
a variant of R Markdown called knitr, which has very similar syntax to R
Markdown for code chunks, can be used in conjunction with the typesetting
system LaTeX to produce documents. Another option is to use the R package
bookdown which uses R Markdown syntax and some additional features to
allow for writing more technical documents. In fact this book was initially
created using knitr and LaTeX, but the simplicity of markdown syntax and
the additional intricacies provided by the bookdown package convinced us to
write the book in R Markdown using bookdown. For simpler tasks, basic R
Markdown is plenty sufficient, and very easy to use.

3.3 Exercises

Exercise 1 Learning objectives: practice setting up a working directory and

read in data; explore the workspace within RStudio and associated commands;
produce basic descriptive statistics and graphics.
Exercise 2 Learning objectives: practice working within RStudio; create a
R Markdown document and resulting html document in RStudio; calculate
descriptive statistics and produce graphics.
4
Data Structures

A data structure is a format for organizing and storing data. The structure
is designed so that data can be accessed and worked with in specific ways.
Statistical software and programming languages have methods (or functions)
designed to operate on different kinds of data structures.
This chapter’s focus is on data structures. To help initial understanding, the
data in this chapter will be relatively modest in size and complexity. The ideas
and methods, however, generalize to larger and more complex data sets.
The base data structures in R are vectors, matrices, arrays, data frames, and
lists. The first three, vectors, matrices, and arrays, require all elements to
be of the same type or homogeneous, e.g., all numeric or all character. Data
frames and lists allow elements to be of different types or heterogeneous, e.g.,
some elements of a data frame may be numeric while other elements may be
character. These base structures can also be organized by their dimensionality,
i.e., 1-dimensional, 2-dimensional, or N-dimensional, as shown in Table 4.1.
R has no scalar types, i.e., 0-dimensional. Individual numbers or strings are
actually vectors of length one.
An efficient way to understand what comprises a given object is to use the
str() function. str() is short for structure and prints a compact, human-
readable description of any R data structure. For example, in the code below,
we prove to ourselves that what we might think of as a scalar value is actually
a vector of length one.

a <- 1
str(a)

TABLE 4.1: Dimension and type content of base data structures in R

Dimension Homogeneous Heterogeneous

1 Atomic vector List
2 Matrix Data frame
N Array

39
40 4 Data Structures

num 1

is.vector(a)

[1] TRUE

length(a)

[1] 1
Here we assigned a the scalar value one. The str(a) prints num 1, which says a
is numeric of length one. Then just to be sure we used the function is.vector()
to test if a is in fact a vector. Then, just for fun, we asked the length of a,
which again returns one. There are a set of similar logical tests for the other
base data structures, e.g., is.matrix(), is.array(), is.data.frame(), and
is.list(). These will all come in handy as we encounter different R objects.

4.1 Vectors
Think of a vector1 as a structure to represent one variable in a data set. For
example a vector might hold the weights, in pounds, of 7 people in a data set.
Or another vector might hold the genders of those 7 people. The c() function
in R is useful for creating (small) vectors and for modifying existing vectors.
Think of c as standing for “combine”.

weight <- c(123, 157, 205, 199, 223, 140, 105)

weight

[1] 123 157 205 199 223 140 105

gender <- c("female", "female", "male", "female", "male",

"male", "female")
gender

[1] "female" "female" "male" "female" "male"

[6] "male" "female"
1 Technically the objects described in this section are “atomic” vectors (all elements of

the same type), since lists, to be described below, also are actually vectors. This will not be
an important issue, and the shorter term vector will be used for atomic vectors below.
4.1 Vectors 41

Notice that elements of a vector are separated by commas when using the c()
function to create a vector. Also notice that character values are placed inside
quotation marks.
The c() function also can be used to add to an existing vector. For example,
if an eighth male person was included in the data set, and his weight was 194
pounds, the existing vectors could be modified as follows.

weight <- c(weight, 194)

gender <- c(gender, "male")
weight

[1] 123 157 205 199 223 140 105 194

gender

[1] "female" "female" "male" "female" "male"

[6] "male" "female" "male"

4.1.1 Types, Conversion, Coercion

Clearly it is important to distinguish between different types of vectors. For

example, it makes sense to ask R to calculate the mean of the weights stored
in weight, but does not make sense to ask R to compute the mean of the
genders stored in gender. Vectors in R may have one of six different “types”:
character, double, integer, logical, complex, and raw. Only the first four of
these will be of interest below, and the distinction between double and integer
will not be of great import. To illustrate logical vectors, imagine that each
of the eight people in the data set was asked whether he or she was taking
blood pressure medication, and the responses were coded as TRUE if the person
answered yes, and FALSE if the person answered no.

typeof(weight)

[1] "double"

typeof(gender)

[1] "character"

bp <- c(FALSE, TRUE, FALSE, FALSE, TRUE, FALSE, TRUE, FALSE)

bp
42 4 Data Structures

[1] FALSE TRUE FALSE FALSE TRUE FALSE TRUE FALSE

typeof(bp)

[1] "logical"
It may be surprising to see that the variable weight is of double type, even
though its values all are integers. By default R creates a double type vector
when numeric values are given via the c() function.
When it makes sense, it is possible to convert vectors to a different type.
Consider the following examples.

weight.int <- as.integer(weight)

weight.int

[1] 123 157 205 199 223 140 105 194

typeof(weight.int)

[1] "integer"

weight.char <- as.character(weight)

weight.char

[1] "123" "157" "205" "199" "223" "140" "105" "194"

bp.double <- as.double(bp)

bp.double

[1] 0 1 0 0 1 0 1 0

gender.oops <- as.double(gender)

Warning: NAs introduced by coercion

gender.oops

[1] NA NA NA NA NA NA NA NA

sum(bp)
4.1 Vectors 43

[1] 3
The integer version of weight doesn’t look any different, but it is stored
differently, which can be important both for computational efficiency and for
interfacing with other languages such as C++. As noted above, however, we will
not worry about the distinction between integer and double types. Converting
weight to character goes as expected: The character representations of the
numbers replace the numbers themselves. Converting the logical vector bp to
double is pretty straightforward too: FALSE is converted to zero, and TRUE is
converted to one. Now think about converting the character vector gender to
a numeric double vector. It’s not at all clear how to represent “female” and
“male” as numbers. In fact in this case what R does is to create a character
vector, but with each element set to NA, which is the representation of missing
data.2 Finally consider the code sum(bp). Now bp is a logical vector, but when
R sees that we are asking to sum this logical vector, it automatically converts
it to a numerical vector and then adds the zeros and ones representing FALSE
and TRUE.
R also has functions to test whether a vector is of a particular type.

is.double(weight)

[1] TRUE

is.character(weight)

[1] FALSE

is.integer(weight.int)

[1] TRUE

is.logical(bp)

[1] TRUE

4.1.1.1 Coercion

Consider the following examples.

2 Missing data will be discussed in more detail later in the chapter.

44 4 Data Structures

xx <- c(1, 2, 3, TRUE)

xx

[1] 1 2 3 1

yy <- c(1, 2, 3, "dog")

yy

[1] "1" "2" "3" "dog"

zz <- c(TRUE, FALSE, "cat")

zz

[1] "TRUE" "FALSE" "cat"

weight+bp

[1] 123 158 205 199 224 140 106 194

Vectors in R can only contain elements of one type. If more than one type is
included in a c() function, R silently coerces the vector to be of one type. The
examples illustrate the hierarchy—if any element is a character, then the whole
vector is character. If some elements are numeric (either integer or double) and
other elements are logical, the whole vector is numeric. Note what happened
when R was asked to add the numeric vector weight to the logical vector bp.
The logical vector was silently coerced to be numeric, so that FALSE became
zero and TRUE became one, and then the two numeric vectors were added.

4.1.2 Accessing Specific Elements of Vectors

To access and possibly change specific elements of vectors, refer to the position
of the element in square brackets. For example, weight[4] refers to the fourth
element of the vector weight. Note that R starts the numbering of elements
at 1, i.e., the first element of a vector x is x[1].

weight

[1] 123 157 205 199 223 140 105 194

weight[5]
4.1 Vectors 45

[1] 223

weight[1:3]

[1] 123 157 205

length(weight)

[1] 8

weight[length(weight)]

[1] 194

weight[]

[1] 123 157 205 199 223 140 105 194

weight[3] <- 202

weight

[1] 123 157 202 199 223 140 105 194

Note that including nothing in the square brackets results in the whole vector
being returned.
Negative numbers in the square brackets tell R to omit the corresponding
value. And a zero as a subscript returns nothing (more precisely, it returns a
length zero vector of the appropriate type).

weight[-3]

[1] 123 157 199 223 140 105 194

weight[-length(weight)]

[1] 123 157 202 199 223 140 105

lessWeight <- weight[-c(1,3,5)]

lessWeight

[1] 157 199 140 105 194

46 4 Data Structures

weight[0]

numeric(0)

weight[c(0,2,1)]

[1] 157 123

weight[c(-1, 2)]

Error in weight[c(-1, 2)]: only 0's may be mixed with negative subscripts
Note that mixing zero and other nonzero subscripts is allowed, but mixing
negative and positive subscripts is not allowed.
What about the (usual) case where we don’t know the positions of the elements
we want? For example possibly we want the weights of all females in the data.
Later we will learn how to subset using logical indices, which is a very powerful
way to access desired elements of a vector.

4.1.3 Practice Problem

A bad programming technique that often plagues beginners is a technique

called hardcoding. Consider the following simple vector containing data on the
number of tree species found at different sites.

tree.sp <- c(10, 13, 15, 8, 2, 9, 10, 20, 9, 11)

Suppose we are interested in the second to last value of the data set. One
way to do this is to first determine the length of vector using the length()
function, then taking that value and subtracting 1.

num.sites <- length(tree.sp)

num.sites

[1] 10

tree.sp[10 - 1]

[1] 9
4.2 Factors 47

This is an example of hardcoding. But what if we attempt to use the same code
on a second vector of tree species data that has a different number of sites?

tree.sp <- c(8, 4, 3, 2, 19, 3)

num.sites <- length(tree.sp)
num.sites

[1] 6

tree.sp[10 - 1]

[1] NA
That’s clearly not what we want. Fix this code so we can always extract the
second to last value in the vector, regardless of the length of the vector.

4.2 Factors
Categorical variables such as gender can be represented as character vectors.
In many cases this simple representation is sufficient. Consider, however, two
other categorical variables, one representing age via categories youth, young
adult, middle age, senior, and another representing income via categories
lower, middle, and upper. Suppose that for the small health data set, all the
people are either middle aged or senior citizens. If we just represented the
variable via a character vector, there would be no way to know that there
are two other categories, representing youth and young adults, which happen
not to be present in the data set. And for the income variable, the character
vector representation does not explicitly indicate that there is an ordering of
the levels.
Factors in R provide a more sophisticated way to represent categorical variables.
Factors explicitly contain all possible levels, and allow ordering of levels.

age <- c("middle age", "senior", "middle age", "senior",

"senior", "senior", "senior", "middle age")
income <- c("lower", "lower", "upper", "middle", "upper",
"lower", "lower", "middle")
age

[1] "middle age" "senior" "middle age" "senior"

[5] "senior" "senior" "senior" "middle age"
48 4 Data Structures

income

[1] "lower" "lower" "upper" "middle" "upper"

[6] "lower" "lower" "middle"

age <- factor(age, levels=c("youth", "young adult", "middle age",

"senior"))
age

[1] middle age senior middle age senior

[5] senior senior senior middle age
Levels: youth young adult middle age senior

income <- factor(income, levels=c("lower", "middle", "upper"),

ordered = TRUE)
income

[1] lower lower upper middle upper lower lower

[8] middle
Levels: lower < middle < upper
In the factor version of age the levels are explicitly listed, so it is clear that
the two included levels are not all the possible levels. And in the factor version
of income, the ordering is explicit.
In many cases the character vector representation of a categorical variable
is sufficient and easier to work with. In this book, factors will not be used
extensively.

4.3 Names of Objects in R

There are few hard and fast restrictions on the names of objects (such as
vectors) in R. In addition to these restrictions, there are certain good practices,
and many things to avoid as well.
From the help page for make.names in R, the name of an R object is “syntacti-
cally valid” if the name “consists of letters, numbers and the dot or underline
characters and starts with a letter or the dot not followed by a number” and
is not one of the “reserved words” in R such as if, TRUE, function, etc. For
example, c45t.le_dog and .ty56 are both syntactically valid (although not
very good names) while 4cats and log#@gopher are not.
4.4 Missing Data, Infinity, etc. 49

A few important comments about naming objects follow:

1. It is important to be aware that names of objects in R are case-

sensitive, so weight and Weight do not refer to the same object.

weight

[1] 123 157 202 199 223 140 105 194

Weight

Error in eval(expr, envir, enclos): object 'Weight' not found

2. It is unwise to create an object with the same name as a built in R

object such as the function c or the function mean. In earlier versions
of R this could be somewhat disastrous, but even in current versions,
it is definitely not a good idea!
3. As much as possible, choose names that are informative. When
creating a variable you may initially remember that x contains
heights and y contains genders, but after a few hours, a few days,
or a few weeks, you probably will forget this. Better options are
Height and Gender.
4. As much as possible, be consistent in how you name objects. In
particular, decide how to separate multi-word names. Some options
include:
•Using case to separate: BloodPressure or bloodPressure for
example
•Using underscores to separate: blood_pressure for example
•Using a period to separate: blood.pressure for example

4.4 Missing Data, Infinity, etc.

Most real-world data sets have variables where some observations are missing.
In a longitudinal study participants may drop out. In a survey, participants
may decide not to respond to certain questions. Statistical software should be
able to represent missing data and to analyze data sets in which some data
are missing.
In R, the value NA is used for a missing data value. Since missing values may
occur in numeric, character, and other types of data, and since R requires
50 4 Data Structures

that a vector contain only elements of one type, there are different types of NA
values. Usually R determines the appropriate type of NA value automatically.
It is worth noting that the default type for NA is logical, and that NA is NOT
the same as the character string "NA".

missingCharacter <- c("dog", "cat", NA, "pig", NA, "horse")

missingCharacter

[1] "dog" "cat" NA "pig" NA "horse"

is.na(missingCharacter)

[1] FALSE FALSE TRUE FALSE TRUE FALSE

missingCharacter <- c(missingCharacter, "NA")

missingCharacter

[1] "dog" "cat" NA "pig" NA "horse"

[7] "NA"

is.na(missingCharacter)

[1] FALSE FALSE TRUE FALSE TRUE FALSE FALSE

allMissing <- c(NA, NA, NA)

typeof(allMissing)

[1] "logical"
How should missing data be treated in computations, such as finding the mean
or standard deviation of a variable? One possibility is to return NA. Another is
to remove the missing value(s) and then perform the computation.

> mean(c(1,2,3,NA,5))

[1] NA

> mean(c(1,2,3,NA,5), na.rm=TRUE)

[1] 2.75
4.4 Missing Data, Infinity, etc. 51

As this example shows, the default behavior for the mean() function is to
return NA. If removal of the missing values and then computing the mean is
desired, the argument na.rm is set to TRUE. Different R functions have different
default behaviors, and there are other possible actions. Consulting the help for
a function provides the details.

4.4.1 Practice Problem

Collecting data is often a messy process resulting in multiple errors in the data.
Consider the following small vector representing the weights of 10 adults in
pounds.

my.weights <- c(150, 138, 289, 239, 12, 103, 310, 200, 218, 178)

As far as I know, it’s not possible for an adult to weigh 12 pounds, so that
is most likely an error. Change this value to NA, and then find the standard
deviation of the weights after removing the NA value.

4.4.2 Infinity and NaN

What happens if R code requests division by zero, or results in a number that

is too large to be represented? Here are some examples.

> x <- 0:4

> x

[1] 0 1 2 3 4

> 1/x

[1] Inf 1.0000 0.5000 0.3333 0.2500

> x/x

[1] NaN 1 1 1 1

> y <- c(10, 1000, 10000)

> 2ˆy

[1] 1.024e+03 1.072e+301 Inf

52 4 Data Structures

Inf and -Inf represent infinity and negative infinity (and numbers which
are too large in magnitude to be represented as floating point numbers). NaN
represents the result of a calculation where the result is undefined, such as
dividing zero by zero. All of these are common to a variety of programming
languages, including R.

4.5 Data Frames

Commonly, data is rectangular in form, with variables as columns and cases
as rows. Continuing with the (contrived) data on weight, gender, and blood
pressure medication, each of those variables would be a column of the data
set, and each person’s measurements would be a row. In R, such data are
represented as a data frame.

healthData <- data.frame(Weight = weight, Gender=gender, bp.meds = bp,

stringsAsFactors=FALSE)
healthData

Weight Gender bp.meds

1 123 female FALSE
2 157 female TRUE
3 202 male FALSE
4 199 female FALSE
5 223 male TRUE
6 140 male FALSE
7 105 female TRUE
8 194 male FALSE

names(healthData)

[1] "Weight" "Gender" "bp.meds"

colnames(healthData)

[1] "Weight" "Gender" "bp.meds"

names(healthData) <- c("Wt", "Gdr", "bp")

healthData
4.5 Data Frames 53

Wt Gdr bp
1 123 female FALSE
2 157 female TRUE
3 202 male FALSE
4 199 female FALSE
5 223 male TRUE
6 140 male FALSE
7 105 female TRUE
8 194 male FALSE

rownames(healthData)

[1] "1" "2" "3" "4" "5" "6" "7" "8"

names(healthData) <- c("Weight", "Gender", "bp.meds")

The data.frame function can be used to create a data frame (although it’s
more common to read a data frame into R from an external file, something that
will be introduced later). The names of the variables in the data frame are given
as arguments, as are the vectors of data that make up the variable’s values. The
argument stringsAsFactors=FALSE asks R not to convert character vectors
into factors. As of version R 4.0.0, R does not automatically convert character
vectors into factors. However, up until this recent version, R would automatically
convert strings to factors (i.e., stringsAsFactors = TRUE), and so to avoid
confusion we will typically display stringsAsFactors=FALSE throughout most
of the book. Names of the columns (variables) can be extracted and set via
either names or colnames. In the example, the variable names are changed
to Wt, Gdr, bp and then changed back to the original Weight, Gender,
bp.meds in this way. Rows can be named also. In this case since specific row
names were not provided, the default row names of "1", "2" etc. are used.
In the next example a built-in dataset called mtcars is made available by the
data function, and then the first and last six rows are displayed using head
and tail.

data(mtcars)
head(mtcars)

mpg cyl disp hp drat wt qsec

Mazda RX4 21.0 6 160 110 3.90 2.620 16.46
Mazda RX4 Wag 21.0 6 160 110 3.90 2.875 17.02
Datsun 710 22.8 4 108 93 3.85 2.320 18.61
Hornet 4 Drive 21.4 6 258 110 3.08 3.215 19.44
Hornet Sportabout 18.7 8 360 175 3.15 3.440 17.02
54 4 Data Structures

Valiant 18.1 6 225 105 2.76 3.460 20.22

vs am gear carb
Mazda RX4 0 1 4 4
Mazda RX4 Wag 0 1 4 4
Datsun 710 1 1 4 1
Hornet 4 Drive 1 0 3 1
Hornet Sportabout 0 0 3 2
Valiant 1 0 3 1

tail(mtcars)

mpg cyl disp hp drat wt qsec vs

Porsche 914-2 26.0 4 120.3 91 4.43 2.140 16.7 0
Lotus Europa 30.4 4 95.1 113 3.77 1.513 16.9 1
Ford Pantera L 15.8 8 351.0 264 4.22 3.170 14.5 0
Ferrari Dino 19.7 6 145.0 175 3.62 2.770 15.5 0
Maserati Bora 15.0 8 301.0 335 3.54 3.570 14.6 0
Volvo 142E 21.4 4 121.0 109 4.11 2.780 18.6 1
am gear carb
Porsche 914-2 1 5 2
Lotus Europa 1 5 2
Ford Pantera L 1 5 4
Ferrari Dino 1 5 6
Maserati Bora 1 5 8
Volvo 142E 1 4 2
Note that the mtcars data frame does have non-default row names which give
the make and model of the cars.

4.5.1 Accessing Specific Elements of Data Frames

Data frames are two-dimensional, so to access a specific element (or elements)

we need to specify both the row and column.

mtcars[1,4]

[1] 110

mtcars[1:3, 3]

[1] 160 160 108

4.5 Data Frames 55

mtcars[1:3, 2:3]

cyl disp
Mazda RX4 6 160
Mazda RX4 Wag 6 160
Datsun 710 4 108

mtcars[,1]

[1] 21.0 21.0 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2
[11] 17.8 16.4 17.3 15.2 10.4 10.4 14.7 32.4 30.4 33.9
[21] 21.5 15.5 15.2 13.3 19.2 27.3 26.0 30.4 15.8 19.7
[31] 15.0 21.4
Note that mtcars[,1] returns ALL elements in the first column. This agrees
with the behavior for vectors, where leaving a subscript out of the square
brackets tells R to return all values. In this case we are telling R to return all
rows, and the first column.
For a data frame there is another way to access specific columns, using the $
notation.

> mtcars$mpg

[1] 21.0 21.0 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2
[11] 17.8 16.4 17.3 15.2 10.4 10.4 14.7 32.4 30.4 33.9
[21] 21.5 15.5 15.2 13.3 19.2 27.3 26.0 30.4 15.8 19.7
[31] 15.0 21.4

> mtcars$cyl

[1] 6 6 4 6 8 6 8 4 4 6 6 8 8 8 8 8 8 4 4 4 4 8 8 8 8
[26] 4 4 4 8 6 8 4

> mpg

Error in eval(expr, envir, enclos): object 'mpg' not found

> cyl

Error in eval(expr, envir, enclos): object 'cyl' not found

 56 4 Data Structures

> weight

[1] 123 157 202 199 223 140 105 194

Notice that typing the variable name, such as mpg, without the name of the
data frame (and a dollar sign) as a prefix, does not work. This is sensible. There
may be several data frames that have variables named mpg, and just typing
mpg doesn’t provide enough information to know which is desired. But if there
is a vector named mpg that is created outside a data frame, it will be retrieved
when mpg is typed, which is why typing weight does work, since weight was
created outside of a data frame, although ultimately it was incorporated into
the healthData data frame.

4.6 Lists
The third main data structure we will work with is a list. Technically a list is
a vector, but one in which elements can be of different types. For example a
list may have one element that is a vector, one element that is a data frame,
and another element that is a function. Consider designing a function that fits
a simple linear regression model to two quantitative variables. We might want
that function to compute and return several things such as
• The fitted slope and intercept (a numeric vector with two components)
• The residuals (a numeric vector with n components, where n is the number
of data points)
• Fitted values for the data (a numeric vector with n components, where n is
the number of data points)
• The names of the dependent and independent variables (a character vector
with two components)
In fact R has a function, lm, which does this (and much more).

mpgHpLinMod <- lm(mpg ~ hp, data=mtcars)

mode(mpgHpLinMod)

[1] "list"

names(mpgHpLinMod)

[1] "coefficients" "residuals" "effects"

4.6 Lists 57

[4] "rank" "fitted.values" "assign"

[7] "qr" "df.residual" "xlevels"
[10] "call" "terms" "model"

mpgHpLinMod$coefficients

(Intercept) hp
30.09886 -0.06823

mpgHpLinMod$residuals

Mazda RX4 Mazda RX4 Wag

-1.59375 -1.59375
Datsun 710 Hornet 4 Drive
-0.95363 -1.19375
Hornet Sportabout Valiant
0.54109 -4.83489
Duster 360 Merc 240D
0.91707 -1.46871
Merc 230 Merc 280
-0.81717 -2.50678
Merc 280C Merc 450SE
-3.90678 -1.41777
Merc 450SL Merc 450SLC
-0.51777 -2.61777
Cadillac Fleetwood Lincoln Continental
-5.71206 -5.02978
Chrysler Imperial Fiat 128
0.29364 6.80421
Honda Civic Toyota Corolla
3.84901 8.23598
Toyota Corona Dodge Challenger
-1.98072 -4.36462
AMC Javelin Camaro Z28
-4.66462 -0.08293
Pontiac Firebird Fiat X1-9
1.04109 1.70421
Porsche 914-2 Lotus Europa
2.10991 8.01093
Ford Pantera L Ferrari Dino
3.71340 1.54109
Maserati Bora Volvo 142E
7.75761 -1.26198
The lm function returns a list (which in the code above has been assigned to
58 4 Data Structures

the object mpgHpLinMod).3 One component of the list is the length 2 vector of
coefficients, while another component is the length 32 vector of residuals. The
code also illustrates that named components of a list can be accessed using
the dollar sign notation, as with data frames.
The list function is used to create lists.

temporaryList <- list(first=weight, second=healthData,

pickle=list(a = 1:10, b=healthData))
temporaryList

$first
[1] 123 157 202 199 223 140 105 194

$second
Weight Gender bp.meds
1 123 female FALSE
2 157 female TRUE
3 202 male FALSE
4 199 female FALSE
5 223 male TRUE
6 140 male FALSE
7 105 female TRUE
8 194 male FALSE

$pickle
$pickle$a
[1] 1 2 3 4 5 6 7 8 9 10

$pickle$b
Weight Gender bp.meds
1 123 female FALSE
2 157 female TRUE
3 202 male FALSE
4 199 female FALSE
5 223 male TRUE
6 140 male FALSE
7 105 female TRUE
8 194 male FALSE
Here, for illustration, I assembled a list to hold some of the R data structures
we have been working with in this chapter. The first list element, named
first, holds the weight vector we created in Section 4.1, the second list
element, named second, holds the healthData data frame, and the third list
3 The mode function returns the type or storage mode of an object.
4.6 Lists 59

element, named pickle, holds a list with elements named a and b that hold a
vector of values 1 through 10 and another copy of the healthData data frame,
respectively. As this example shows, a list can contain another list.

4.6.1 Accessing Specific Elements of Lists

We already have seen the dollar sign notation works for lists. In addition, the
square bracket subsetting notation can be used. There is an added, somewhat
subtle wrinkle—using either single or double square brackets.

temporaryList$first

[1] 123 157 202 199 223 140 105 194

mode(temporaryList$first)

[1] "numeric"

temporaryList[[1]]

[1] 123 157 202 199 223 140 105 194

mode(temporaryList[[1]])

[1] "numeric"

temporaryList[1]

$first
[1] 123 157 202 199 223 140 105 194

mode(temporaryList[1])

[1] "list"
Note the dollar sign and double bracket notation return a numeric vector,
while the single bracket notation returns a list. Notice also the difference in
results below.

temporaryList[c(1,2)]
60 4 Data Structures

$first
[1] 123 157 202 199 223 140 105 194

$second
Weight Gender bp.meds
1 123 female FALSE
2 157 female TRUE
3 202 male FALSE
4 199 female FALSE
5 223 male TRUE
6 140 male FALSE
7 105 female TRUE
8 194 male FALSE

temporaryList[[c(1,2)]]

[1] 157
The single bracket form returns the first and second elements of the list, while
the double bracket form returns the second element in the first element of the
list. Generally, do not put a vector of indices or names in a double bracket,
you will likely get unexpected results. See, for example, the results below.4

temporaryList[[c(1,2,3)]]

Error in temporaryList[[c(1, 2, 3)]]: recursive indexing failed at level 2

So, in summary, there are two main differences between using the single bracket
[] and double bracket [[]]. First, the single bracket will return a list that
holds the object(s) held at the given indices or names placed in the bracket,
whereas the double brackets will return the actual object held at the index
or name placed in the innermost bracket. Put differently, a single bracket can
be used to access a range of list elements and will return a list, and a double
bracket can only access a single element in the list and will return the object
held at the index.
4 Try this example using only single brackets. . . it will return a list holding elements

first, second, and pickle.

4.7 Subsetting with Logical Vectors 61

4.7 Subsetting with Logical Vectors

Consider the healthData data frame. How can we access only those weights
which are more than 200? How can we access the genders of those whose
weights are more than 200? How can we compute the mean weight of males
and the mean weight of females? Or consider the mtcars data frame. How can
we obtain the miles per gallon for all six cylinder cars? Both of these data sets
are small enough that it would not be too onerous to extract the values by
hand. But for larger or more complex data sets, this would be very difficult or
impossible to do in a reasonable amount of time, and would likely result in
errors.
R has a powerful method for solving these sorts of problems using a variant of
the subsetting methods that we already have learned. When given a logical
vector in square brackets, R will return the values corresponding to TRUE. To
begin, focus on the weight and gender vectors created in Section 4.1.
The R code weight > 200 returns a TRUE for each value of weight which is
more than 200, and a FALSE for each value of weight which is less than or
equal to 200. Similarly gender == "female" returns TRUE or FALSE depending
on whether an element of gender is equal to female.

weight

[1] 123 157 202 199 223 140 105 194

weight > 200

[1] FALSE FALSE TRUE FALSE TRUE FALSE FALSE FALSE

gender[weight > 200]

[1] "male" "male"

weight[weight > 200]

[1] 202 223

gender == "female"

[1] TRUE TRUE FALSE TRUE FALSE FALSE TRUE FALSE

 62 4 Data Structures

weight[gender == "female"]

[1] 123 157 199 105

Consider the lines of R code one by one.
• weight instructs R to display the values in the vector weight.
• weight > 200 instructs R to check whether each value in weight is greater
than 200, and to return TRUE if so, and FALSE otherwise.
• The next line, gender[weight > 200], does two things. First, inside the
square brackets, it does the same thing as the second line, namely, returning
TRUE or FALSE depending on whether a value of weight is or is not greater
than 200. Second, each element of gender is matched with the corresponding
TRUE or FALSE value, and is returned if and only if the corresponding value
is TRUE. For example the first value of gender is gender[1]. Since the first
TRUE or FALSE value is FALSE, the first value of gender is not returned. Only
the third and fifth values of gender, both of which happen to be male, are
returned. Briefly, this line returns the genders of those people whose weight
is over 200 pounds.
• The fourth line of code, weight[weight > 200], again begins by returning
TRUE or FALSE depending on whether elements of weight are larger than 200.
Then those elements of weight corresponding to TRUE values, are returned.
So this line returns the weights of those people whose weights are more than
200 pounds.
• The fifth line returns TRUE or FALSE depending on whether elements of
gender are equal to female or not.
• The sixth line returns the weights of those whose gender is female.
There are six comparison operators in R, >, <, >=, <=, ==, !=. Note that
to test for equality a “double equals sign” is used, while != tests for inequality.

4.7.1 Modifying or Creating Objects via Subsetting

The results of subsetting can be assigned to a new (or existing) R object, and
subsetting on the left side of an assignment is a common way to modify an
existing R object.

weight

[1] 123 157 202 199 223 140 105 194

lightweight <- weight[weight < 200]

lightweight
4.7 Subsetting with Logical Vectors 63

[1] 123 157 199 140 105 194

x <- 1:10
x

[1] 1 2 3 4 5 6 7 8 9 10

x[x < 5] <- 0

x

[1] 0 0 0 0 5 6 7 8 9 10

y <- -3:9
y

[1] -3 -2 -1 0 1 2 3 4 5 6 7 8 9

y[y < 0] <- NA

y

[1] NA NA NA 0 1 2 3 4 5 6 7 8 9

rm(x)
rm(y)

4.7.2 Logical Subsetting and Data Frames

First consider the small and simple healthData data frame.

healthData

Weight Gender bp.meds

1 123 female FALSE
2 157 female TRUE
3 202 male FALSE
4 199 female FALSE
5 223 male TRUE
6 140 male FALSE
7 105 female TRUE
8 194 male FALSE
64 4 Data Structures

healthData$Weight[healthData$Gender == "male"]

[1] 202 223 140 194

healthData[healthData$Gender == "female", ]

Weight Gender bp.meds

1 123 female FALSE
2 157 female TRUE
4 199 female FALSE
7 105 female TRUE

healthData[healthData$Weight > 190, 2:3]

Gender bp.meds
3 male FALSE
4 female FALSE
5 male TRUE
8 male FALSE
The first example is really just subsetting a vector, since the $ notation creates
vectors. The second two examples return subsets of the whole data frame. Note
that the logical vector subsets the rows of the data frame, choosing those rows
where the gender is female or the weight is more than 190. Note also that the
specification for the columns (after the comma) is left blank in the first case,
telling R to return all the columns. In the second case the second and third
columns are requested explicitly.
Next consider the much larger and more complex WorldBank data frame.
Recall, the str function displays the “structure” of an R object. Here is a look
at the structure of several R objects.

str(mtcars)

'data.frame': 32 obs. of 11 variables:

$ mpg : num 21 21 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2 ...
$ cyl : num 6 6 4 6 8 6 8 4 4 6 ...
$ disp: num 160 160 108 258 360 ...
$ hp : num 110 110 93 110 175 105 245 62 95 123 ...
$ drat: num 3.9 3.9 3.85 3.08 3.15 2.76 3.21 3.69 3.92 3.92 ...
$ wt : num 2.62 2.88 2.32 3.21 3.44 ...
$ qsec: num 16.5 17 18.6 19.4 17 ...
$ vs : num 0 0 1 1 0 1 0 1 1 1 ...
4.7 Subsetting with Logical Vectors 65

$ am : num 1 1 1 0 0 0 0 0 0 0 ...
$ gear: num 4 4 4 3 3 3 3 4 4 4 ...
$ carb: num 4 4 1 1 2 1 4 2 2 4 ...

str(temporaryList)

List of 3
$ first : num [1:8] 123 157 202 199 223 140 105 194
$ second:'data.frame': 8 obs. of 3 variables:
..$ Weight : num [1:8] 123 157 202 199 223 140 105 194
..$ Gender : chr [1:8] "female" "female" "male" "female" ...
..$ bp.meds: logi [1:8] FALSE TRUE FALSE FALSE TRUE FALSE ...
$ pickle:List of 2
..$ a: int [1:10] 1 2 3 4 5 6 7 8 9 10
..$ b:'data.frame': 8 obs. of 3 variables:
.. ..$ Weight : num [1:8] 123 157 202 199 223 140 105 194
.. ..$ Gender : chr [1:8] "female" "female" "male" "female" ...
.. ..$ bp.meds: logi [1:8] FALSE TRUE FALSE FALSE TRUE FALSE ...

str(WorldBank)

'data.frame': 11880 obs. of 15 variables:

$ iso2c : chr "AD" "AD" "AD" "AD" ...
$ country : chr "Andorra" "Andorra" "Andorra" "Andorra" ...
$ year : int 1978 1979 1977 2007 1976 2011 2012 2008 1980 1972 ..
$ fertility.rate : num NA NA NA 1.18 NA NA NA 1.25 NA NA ...
$ life.expectancy : num NA NA NA NA NA NA NA NA NA NA ...
$ population : num 33746 34819 32769 81292 31781 ...
$ GDP.per.capita.Current.USD : num 9128 11820 7751 39923 7152 ...
$ X15.to.25.yr.female.literacy: num NA NA NA NA NA NA NA NA NA NA ...
$ iso3c : chr "AND" "AND" "AND" "AND" ...
$ region : chr "Europe & Central Asia (all income levels)" "Europe
$ capital : chr "Andorra la Vella" "Andorra la Vella" "Andorra la Ve
$ longitude : num 1.52 1.52 1.52 1.52 1.52 ...
$ latitude : num 42.5 42.5 42.5 42.5 42.5 ...
$ income : chr "High income: nonOECD" "High income: nonOECD" "High
$ lending : chr "Not classified" "Not classified" "Not classified" "
First we see that mtcars is a data frame which has 32 observations (rows) on
each of 11 variables (columns). The names of the variables are given, along
with their type (in this case, all numeric), and the first few values of each
variable is given.
Second we see that temporaryList is a list with three components. Each of
the components is described separately, with the first few values again given.
66 4 Data Structures

Third we examine the structure of WorldBank. It is a data frame with 11880

observations on each of 15 variables. Some of these are character variables,
some are numeric, and one (year) is integer. Looking at the first few values
we see that some variables have missing values.
Consider creating a data frame which only has the observations from one year,
say 1971. That’s relatively easy. Just choose rows for which year is equal to
1971.

WorldBank1971 <- WorldBank[WorldBank$year == 1971, ]

dim(WorldBank1971)

[1] 216 15
The dim function returns the dimensions of a data frame, i.e., the num-
ber of rows and the number of columns. From dim we see that there are
dim(WorldBank1971)[1] cases from 1971.
Next, how can we create a data frame which only contains data from 1971,
and also only contains cases for which there are no missing values in the
fertility rate variable? R has a built in function is.na which returns TRUE if
the observation is missing and returns FALSE otherwise. And !is.na returns
the negation, i.e., it returns FALSE if the observation is missing and TRUE if
the observation is not missing.

WorldBank1971$fertility.rate[1:25]

[1] NA 6.512 7.671 3.517 4.933 3.118 7.264 3.104

[9] NA 2.200 2.961 2.788 4.479 2.260 2.775 2.949
[17] 6.942 2.210 6.657 2.100 6.293 7.329 6.786 NA
[25] 5.771

!is.na(WorldBank1971$fertility.rate[1:25])

[1] FALSE TRUE TRUE TRUE TRUE TRUE TRUE TRUE

[9] FALSE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
[17] TRUE TRUE TRUE TRUE TRUE TRUE TRUE FALSE
[25] TRUE

WorldBank1971 <- WorldBank1971[!is.na(WorldBank1971$fertility.rate),]

dim(WorldBank1971)

[1] 193 15
4.7 Subsetting with Logical Vectors 67

From dim we see that there are 193 cases from 1971 with non-missing fertility
rate data.
Return attention now to the original WorldBank data frame with data not
only from 1971. How can we extract only those cases (rows) which have NO
missing data? Consider the following simple example:

temporaryDataFrame <- data.frame(V1 = c(1, 2, 3, 4, NA),

V2 = c(NA, 1, 4, 5, NA),
V3 = c(1, 2, 3, 5, 7))
temporaryDataFrame

V1 V2 V3
1 1 NA 1
2 2 1 2
3 3 4 3
4 4 5 5
5 NA NA 7

is.na(temporaryDataFrame)

V1 V2 V3
[1,] FALSE TRUE FALSE
[2,] FALSE FALSE FALSE
[3,] FALSE FALSE FALSE
[4,] FALSE FALSE FALSE
[5,] TRUE TRUE FALSE

rowSums(is.na(temporaryDataFrame))

[1] 1 0 0 0 2
First notice that is.na will test each element of a data frame for missingness.
Also recall that if R is asked to sum a logical vector, it will first convert the
logical vector to numeric and then compute the sum, which effectively counts the
number of elements in the logical vector which are TRUE. The rowSums function
computes the sum of each row. So rowSums(is.na(temporaryDataFrame))
returns a vector with as many elements as there are rows in the data frame. If
an element is zero, the corresponding row has no missing values. If an element
is greater than zero, the value is the number of variables which are missing in
that row. This gives a simple method to return all the cases which have no
missing data.
68 4 Data Structures

dim(WorldBank)

[1] 11880 15

WorldBankComplete <- WorldBank[rowSums(is.na(WorldBank)) == 0,]

dim(WorldBankComplete)

[1] 564 15
Out of the 564 rows in the original data frame, only 564 have no missing
observations!

4.8 Patterned Data

Sometimes it is useful to generate all the integers from 1 through 20, to
generate a sequence of 100 points equally spaced between 0 and 1, etc. The R
functions seq() and rep() as well as the “colon operator” : help to generate
such sequences.
The colon operator generates a sequence of values with increments of 1 or −1.

1:10

[1] 1 2 3 4 5 6 7 8 9 10

-5:3

[1] -5 -4 -3 -2 -1 0 1 2 3

10:4

[1] 10 9 8 7 6 5 4

pi:7

[1] 3.142 4.142 5.142 6.142

The seq() function generates either a sequence of pre-specified length or a
sequence with pre-specified increments.
4.8 Patterned Data 69

seq(from = 0, to = 1, length = 11)

[1] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

seq(from = 1, to = 5, by = 1/3)

[1] 1.000 1.333 1.667 2.000 2.333 2.667 3.000 3.333

[9] 3.667 4.000 4.333 4.667 5.000

seq(from = 3, to = -1, length = 10)

[1] 3.0000 2.5556 2.1111 1.6667 1.2222 0.7778

[7] 0.3333 -0.1111 -0.5556 -1.0000
The rep() function replicates the values in a given vector.

rep(c(1,2,4), length = 9)

[1] 1 2 4 1 2 4 1 2 4

rep(c(1,2,4), times = 3)

[1] 1 2 4 1 2 4 1 2 4

rep(c("a", "b", "c"), times = c(3, 2, 7))

[1] "a" "a" "a" "b" "b" "c" "c" "c" "c" "c" "c" "c"

4.8.1 Practice Problem

Often when using R you will want to simulate data from a specific probability
distribution (i.e. normal/Gaussian, bionmial, Poisson). R has a vast suite
of functions for working with statistical distributions. To generate values
from a statistical distribution, the function has a name beginning with an “r”
followed by some abbreviation of the probability distribution. For example
to simulate from the three distributions mentioned above, we can use the
functions rnorm(), rbinom(), and rpois().
Use the rnorm() function to generate 10,000 values from the standard normal
distribution (the normal distribution with mean = 0 and variance = 1). Consult
the help page for rnorm() if you need to. Save this vector of variables to a
70 4 Data Structures

vector named sim.vals. Then use the hist() function to draw a histogram
of the simulated data. Does the data look like it follows a normal distribution?

4.9 Exercises

Exercise 3 Learning objectives: create, subset, and manipulate vector contents

and attributes; summarize vector data using R table() and other functions;
generate basic graphics using vector data.
Exercise 4 Learning objectives: use functions to describe data frame char-
acteristics; summarize and generate basic graphics for variables held in data
frames; apply the subset function with logical operators; illustrate NA, NaN,
Inf, and other special values; recognize the implications of using floating point
arithmetic with logical operators.
Exercise 5 Learning objectives: practice with lists, data frames, and associated
functions; summarize variables held in lists and data frames; work with R’s
linear regression lm() function output; review logical subsetting of vectors for
partitioning and assigning of new values; generate and visualize data from
mathematical functions.
5
Graphics in R Part 1: ggplot2

R can be used to create a vast array of graphical representations of data.

Creating “standard” graphical displays is straightforward, but a main strength
of R is the ability to customize graphical displays to create either non-standard
graphics or to modify more standard graphical displays to create publication-
ready versions.
There are several packages available in R for creating graphics. The two leading
packages are the graphics package, which comes with your base installation
of R, and the ggplot2 package, which must be installed and made available
by the user.1 For beginners ggplot2 has somewhat simpler syntax, and also
produces excellent graphics without much tinkering. However, the graphics
package seemingly provides more control over different graphical parameters,
and can sometimes be more intuitive than ggplot2.
Knowing how to use both the graphics and ggplot2 packages is worthwhile,
so we denote one chapter to ggplot2 and a second chapter to graphics. We
use ggplot2 throughout the text, and thus require you to read this chapter on
ggplot2 while Chapter 15 on the graphics package is optional. You’ll notice
that the chapters are essentially the same, producing the same graphs (with a
few exceptions) to let you decide which graphing style you prefer. We simply
want to present you with both sets of tools so that in the future when you
have graphs to produce you can use whichever package floats your boat! For
now we’ll start off with ggplot2 and get to graphics in Chapter 15.
The gg in ggplot2 stands for Grammar of Graphics. The package provides
a unified and logical way to describe graphical displays such as scatter plots,
histograms, bar charts, and many other types of graphics. The grammar
describes the mapping from data to the graphical display’s aesthetic attributes
(color, shape, size) of geometric objects (points, lines, bars). As will become
obvious, once this grammar is mastered for a particular type of plot, such as a
scatter plot, it is easy to transfer this knowledge to other types of graphics.
Once you work through this chapter, the best place to learn more about
ggplot2 is from the package’s official book Wickham and Sievert (2016) by
Hadley Wickham. It is available on-line in digital format from MSU’s library.
The book goes into much more depth on the theory underlying the grammar
1 Other graphics packages include lattice and grid

71
72 5 Graphics in R Part 1: ggplot2

and syntax, and has many examples on solving practical graphical problems. In
addition to the free on-line version available through MSU, the book’s source
code is available at https://github.com/hadley/ggplot2-book.

5.1 Scatter Plots

Scatter plots are a workhorse of data visualization and provide a good entry
point to the ggplot2 system. Begin by considering a simple and classic data
set sometimes called Fisher’s Iris Data. These data are available in R.

data(iris)
str(iris)

'data.frame': 150 obs. of 5 variables:

$ Sepal.Length: num 5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
$ Sepal.Width : num 3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
$ Petal.Length: num 1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
$ Petal.Width : num 0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
$ Species : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...
The data contain measurements on petal and sepal length and width for
150 iris plants. The plants are from one of three species, and the species
information is also included in the data frame. The data are commonly used to
test classification methods, where the goal would be to correctly determine the
species based on the four length and width measurements. To get a preliminary
sense of how this might work, we can draw some scatter plots of length versus
width. Recall that ggplot2 is not available by default, so we first have to
download and install the package.

install.packages("ggplot2")

Once this is done the package is installed on the local hard drive, and we can
use the library function to make the package available during the current R
session.
Next a basic scatter plot is drawn. We’ll keep the focus on sepal length and
width, but of course similar plots could be drawn using petal length and width.
The prompt is not displayed below, since the continuation prompt + can cause
confusion.
5.1 Scatter Plots 73

library(ggplot2)
ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +
geom_point()

4.5

4.0

3.5
Sepal.Width

3.0

2.5

2.0

5 6 7 8
Sepal.Length

In this case the first argument to the ggplot function is the name of the data
frame. Second, the aes (short for aesthetics) function specifies the mapping to
the x and y axes. By itself the ggplot function as written doesn’t tell R what
sort of graphical display is desired. That is done by adding a geom (short for
geometry) specification, in this case geom_point.
Looking at the scatter plot and thinking about the focus of finding a method
to classify the species, two thoughts come to mind. First, the plot might be
improved by increasing the size of the points. And second, using different colors
for the points corresponding to the three species would help.

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size = 4, aes(color=Species))
74 5 Graphics in R Part 1: ggplot2
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

Notice that a legend showing what the colors represent is automatically gener-
ated and included in the graphic. Next, the size of the points seems a bit big
now, and it might be helpful to use different shapes for the different species.

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size = 3, aes(color=Species, shape=Species))
5.1 Scatter Plots 75
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

Here we see that the legend automatically changes to include species specific
color and shape. The size of the points seems more appropriate.

5.1.1 Structure of a Typical ggplot

The examples above start with the function ggplot(), which takes as argu-
ments the data frame containing the data to be plotted as well as a mapping
from the data to the axes, enclosed by the aes() function. Next a geom func-
tion, in the above case geom_point(), is added. It might just specify the
geometry, but also might specify aspects such as size, color, or shape.
Typically many graphics are created and discarded in the search for an infor-
mative graphic, and often the initial specification of data and basic aesthetics
from ggplot() stays the same in all the attempts. In such a case it can be
helpful to assign that portion of the graphic to an R object, both to minimize
the amount of typing and to keep certain aspects of all the graphics constant.
Here’s how that could be done for the graphics above.
76 5 Graphics in R Part 1: ggplot2

iris.p <- ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width))

iris.p + geom_point()

4.5

4.0

3.5
Sepal.Width

3.0

2.5

2.0

5 6 7 8
Sepal.Length

iris.p + geom_point(size = 3, aes(color=Species, shape=Species))

5.1 Scatter Plots 77
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

5.1.2 Adding lines to a scatter plot

To add a fitted least squares line to a scatter plot, use stat_smooth, which
adds a smoother (possibly a least squares line, possibly a smooth curve fit
to the data, etc.). The argument method = lm specifies a line fitted by least
squares, and the argument se = FALSE suppresses the default display of a
confidence band around the line or curve which was fit to the data.

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size=3, aes(color=Species))
78 5 Graphics in R Part 1: ggplot2
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size=3, aes(color=Species)) +
stat_smooth(method = lm, se=FALSE)

`geom_smooth()` using formula 'y ~ x'

5.1 Scatter Plots 79
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

For the iris data, it probably makes more sense to fit separate lines by species.
This can be specified using the aes() function inside stat_smooth().

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size=3, aes(color=Species)) +
stat_smooth(method = lm, se=FALSE, aes(color=Species))

`geom_smooth()` using formula 'y ~ x'

80 5 Graphics in R Part 1: ggplot2
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

In this case we specified the same color aesthetic for the points and the lines.
If we know we want this color aesthetic (colors corresponding to species) for
all aspects of the graphic, we can specify it in the main ggplot() function:

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width, color = Species)) +

geom_point(size=3) + stat_smooth(method = lm, se=FALSE)

`geom_smooth()` using formula 'y ~ x'

5.1 Scatter Plots 81
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

Another common use of line segments in a graphic is to connect the points in

order, accomplished via the geom_line() function. Although it is not clear
why this helps in understanding the iris data, the technique is illustrated
next, first doing this for all the points in the graphic, and second doing this
separately for the three species.

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size = 4, aes(color=Species, shape = Species)) +
geom_line()
82 5 Graphics in R Part 1: ggplot2
4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(size = 4, aes(color=Species)) +
geom_line(aes(color=Species))

4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length
5.2 Labels, Axes, Text, etc. 83

5.2 Labels, Axes, Text, etc.

The default settings of ggplot2 often produce excellent graphics, but once a
graphic is chosen for dissemination, the user will likely want to customize things
like the title, axes, etc. In this section some tools for customization are presented.
Most will be illustrated in the context of a data set on crime rates in the 50 states
in the United States. These data were made available by Nathan Yau at http:
//flowingdata.com/2010/11/23/how-to-make-bubble-charts/. The data
include crime rates per 100,000 people for various crimes such as murder and
robbery, and also include each state’s population. The crime rates are from
the year 2005, while the population numbers are from the year 2008, but
the difference in population between the years is not great, and the exact
population is not particularly important for what we’ll do below.
First, read in the data, examine its structure, and produce a simple scatter
plot of motor vehicle theft versus burglary.

u.crime <- "https://www.finley-lab.com/files/data/crimeRatesByState2005.csv"

crime <- read.csv(u.crime, header=TRUE)
str(crime)

'data.frame': 50 obs. of 9 variables:

$ state : chr "Alabama " "Alaska " "Arizona " "Arkansas" ...
$ murder : num 8.2 4.8 7.5 6.7 6.9 3.7 2.9 4.4 5 6.2 ...
$ Forcible_rate : num 34.3 81.1 33.8 42.9 26 43.4 20 44.7 37.1 23.6 ...
$ Robbery : num 141.4 80.9 144.4 91.1 176.1 ...
$ aggravated_assult : num 248 465 327 387 317 ...
$ burglary : num 954 622 948 1085 693 ...
$ larceny_theft : num 2650 2599 2965 2711 1916 ...
$ motor_vehicle_theft: num 288 391 924 262 713 ...
$ population : int 4627851 686293 6500180 2855390 36756666 4861515 3501252 87309

ggplot(data <- crime, aes(x = burglary, y = motor_vehicle_theft)) +

geom_point()
84 5 Graphics in R Part 1: ggplot2

900
motor_vehicle_theft

600

300

500 750 1000

burglary

5.2.1 Labels

By default axis and legend labels are the names of the relevant columns in the
data frame. While convenient, we often want to customize these labels. Here
we use labs() to change the x and y axis labels and other descriptive text.2

ggplot(data = crime, aes(x = burglary, y = motor_vehicle_theft)) +

geom_point() +
labs(x = "Burglaries per 100,000 population",
y = "Motor vehicle theft per 100,000 population",
title = "Burglaries vs motor vehicle theft for US states",
subtitle = "2005 crime rates and 2008 population",
caption = "Data from Nathan Yau http://flowingdata.com"
)

2 Axis and legend labels can also be set in the individual scales, see the subsequent

sections.
5.3 Customizing Axes 85

Burglaries vs motor vehicle theft for US states

2005 crime rates and 2008 population
Motor vehicle theft per 100,000 population

900

600

300

500 750 1000

Burglaries per 100,000 population
Data from Nathan Yau http://flowingdata.com

5.3 Customizing Axes

ggplot also provides default axis extents (i.e., limits) and other axis features.
These, and other axis features such as tick marks, labels, and transformations,
can be changed using the scale functions. Here the range of the x and y axis is
altered to start at zero and go to the maximum of the x and y variables.3 Here
too, axis labels are specified within the scale function, which is an alterative
to using the labs() function.

ggplot(data = crime, aes(x = burglary, y = motor_vehicle_theft)) +

geom_point() +
scale_x_continuous(name="Burglaries per 100,000 population",
limits=c(0,max(crime$burglary))) +
scale_y_continuous(name="Motor vehicle theft per 100,000 population",
limits = c(0, max(crime$motor_vehicle_theft)))

3 ggplot2 makes the axes extend slightly beyond the given range, since typically this is

what the user wants.

86 5 Graphics in R Part 1: ggplot2

Motor vehicle theft per 100,000 population

900

600

300

0 250 500 750 1000 1250

Burglaries per 100,000 population

5.3.1 Text, Point Size, and Color

Next we make point size proportional to population, change the color, and
add a state label. Note, in the ggplot() call I scaled population by 100,000
to help with the interpretability of the legend. Accordingly, I also changed
the “population” label on the legend to “Population\n(100,000)” using the
labs() function4 . We use the geom_label() function to add the label, which
provides an outline around the label text and allows you to control the box
characteristics, e.g., I make the boxes slightly transparent using the alpha
argument.5

ggplot(data = crime, aes(x = burglary, y = motor_vehicle_theft,

size=population/100000)) +
geom_point(color = "blue") +
geom_label(aes(label = state), alpha = 0.5) +
scale_x_continuous(name="Burglaries per 100,000 population",
limits=c(0,max(crime$burglary))) +
scale_y_continuous(name="Motor vehicle theft per 100,000 population",

4 The \n is the line break and puts “(100,000)” below “Population”.

5 You can also add labels via geom_text() function or a label argument in the ggplot()
call.
5.3 Customizing Axes 87

limits = c(0, max(crime$motor_vehicle_theft))) +

labs(size="Population\n(100,000)")

Nevada
Motor vehicle theft per 100,000 population

Arizona
900

Washington

California Hawaii
Population
(100,000)
600 Maryland
Colorado a 100
Oregon
Michigan Georgia a 200

a
Missouri
Rhode Island Florida
Texas
Tennessee
New Mexico
Oklahoma
300
Alaska
South Carolina
Utah Indiana
Kansas
Ohio
New Jersey North Carolina
300 Illinois
Nebraska
Massachusetts Delaware
Connecticut
Minnesota
Louisiana
Alabama
Mississippi Arkansas
PennsylvaniaWest
Wisconsin
Virginia Kentucky
New York
Montana VirginiaIdaho
North Dakota
Iowa
Wyoming

NewSouth Dakota
Hampshire Maine
Vermont

0 250 500 750 1000 1250

Burglaries per 100,000 population

The labels are helpful but just too cluttered. There are some additional
arguments that can go into geom_label() that allow for label offset; however,
this won’t help us much here. Instead, we can try the ggrepel package by
Kamil Slowikowski. This useful package will automatically adjust labels so that
they don’t overlap. First we need to download and add the package using either
RStudio’s install package buttons or via install.packages("ggrepel"). Next
to make all of ggrepel’s functions available we can call library(ggrepel)
function or, if we know which function we want, we can load only that particular
function using the :: operators. I use :: below to make clear which function
is coming from ggrepel and which is coming from ggplot2.

ggplot(data = crime, aes(x = burglary, y = motor_vehicle_theft,

size=population/100000)) +
geom_point(color = "blue") +
scale_x_continuous(name="Burglaries per 100,000 population",
limits=c(0,max(crime$burglary))) +
scale_y_continuous(name="Motor vehicle theft per 100,000 population",
limits = c(0, max(crime$motor_vehicle_theft))) +
labs(size="Population\n(100,000)") +
ggrepel::geom_label_repel(aes(label = state), alpha = 0.5)
88 5 Graphics in R Part 1: ggplot2

Nevada
Motor vehicle theft per 100,000 population

900 Arizona

California Hawaii
Washington

Maryland Georgia Tennessee Population

Colorado
Alaska
(100,000)
600 Oregon
Texas a 100
Rhode Island Michigan
Missouri
New Mexico a 200
New Jersey Utah

Indiana Florida Oklahoma a 300

Connecticut Nebraska
South Carolina
300 Kansas Ohio North Carolina
Montana
Pennsylvania
Minnesota
Louisiana Alabama
New York Wisconsin
Massachusetts Kentucky Arkansas
North Dakota South Dakota Virginia Delaware

Iowa
Illinois
Wyoming
Mississippi
New Hampshire Maine
0 Vermont Idaho West Virginia

0 250 500 750 1000 1250

Burglaries per 100,000 population

This looks a bit better. We’ll resist making additional improvements to the
figure for now.

5.3.2 Practice Problem

You run an experiment to see if the number of alcoholic beverages a person

has on average every day is related to weight. You collect 15 data points. Enter
these data in R by creating vectors, and then reproduce the following plot.
5.4 Customizing Axes 89

# of Drinks Weight
1.0 150
3.0 350
2.0 200
0.5 140
2.0 200
1.0 160
0.0 175
0.0 140
0.0 167
1.0 200
4.0 300
5.0 321
2.0 250
0.5 187
1.0 190

350

300
Weight

250

200

150

0 1 2 3 4 5
Number of Drinks
90 5 Graphics in R Part 1: ggplot2

5.4 Other Types of Graphics

Scatter and line plots, which have just been presented, are common but
certainly not the only graphical displays in common use. Histograms, boxplots,
and bar graphs, as well as more “mathematical” displays such as the graph
of a function, are commonly used to represent data. Examples of each are
presented below.

5.4.1 Histograms

Simon Newcomb conducted several experiments to estimate the speed of light

by measuring the time it took for light to travel from his laboratory to a mirror
at the base of the Washington Monument, and then back to his lab. This is a
distance of 7.44373 km, and by dividing this distance by the measured time,
an estimate for the speed of light is obtained.
The times are of course quite small, and to avoid working with very small
numbers, the data are recoded to be the deviation from 24800 nanoseconds. For
example an observation coded as 28 represents a time of 24828 nanoseconds,
while an observation coded as −44 represents a time of 24756 nanoseconds.

u.newcomb <- "https://www.finley-lab.com/files/data/Newcomb.csv"

Newcomb <- read.csv(u.newcomb, header=TRUE)
head(Newcomb)

Time
1 28
2 26
3 33
4 24
5 34
6 -44

ggplot(Newcomb, aes(x = Time)) + geom_histogram()

`stat_bin()` using `bins = 30`. Pick better value

with `binwidth`.
5.4 Other Types of Graphics 91

15

10
count

-25 0 25
Time

The software has an algorithm to calculate bin widths for the histogram.
Sometimes the algorithm makes choices that aren’t suitable (hence the R
message above), and these can be changed by specifying a binwidth. In
addition, the appearance of the bars also can be changed.

ggplot(Newcomb, aes(x = Time)) +

geom_histogram(binwidth = 5, color = "black", fill = "blue" )
92 5 Graphics in R Part 1: ggplot2
25

20

15
count

10

-50 -25 0 25
Time

5.4.2 Boxplots

Next we consider some data from the gap minder data set to construct some
box plots. These data are available in the gapminder package, which might
need to be installed via install.packages("gapminder").

library(gapminder)
ggplot(data = subset(gapminder, year == 2002),
aes(x = continent, y = gdpPercap)) +
geom_boxplot(color = "black", fill = "lightblue")
5.4 Other Types of Graphics 93

40000

30000
gdpPercap

20000

10000

Africa Americas Asia Europe Oceania

continent

Here’s the same set of boxplots, but with different colors, different axis labels,
and the boxes plotted horizontally rather than vertically.

ggplot(data = subset(gapminder, year == 2002),

aes(x = continent, y = gdpPercap)) +
geom_boxplot(color = "red", fill = "lightblue") +
scale_x_discrete(name = "Continent") +
scale_y_continuous(name = "Per Capita GDP") + coord_flip()
94 5 Graphics in R Part 1: ggplot2

Oceania

Europe
Continent

Asia

Americas

Africa

0 10000 20000 30000 40000

Per Capita GDP

5.4.3 Bar Graphs

As part of a study, elementary school students were asked which was more
important to them: good grades, popularity, or athletic ability. Here is a brief
look at the data.

u.goals <- "https://www.finley-lab.com/files/data/StudentGoals.csv"

StudentGoals <- read.csv(u.goals, header=TRUE)
head(StudentGoals)

Gender Grade Age Race Type School Goals Grades

1 boy 5 11 White Rural Elm Sports 1
2 boy 5 10 White Rural Elm Popular 2
3 girl 5 11 White Rural Elm Popular 4
4 girl 5 11 White Rural Elm Popular 2
5 girl 5 10 White Rural Elm Popular 4
6 girl 5 11 White Rural Elm Popular 4
Sports Looks Money
1 2 4 3
2 1 4 3
3 3 1 2
4 3 4 1
5 2 1 3
5.4 Other Types of Graphics 95

6 2 1 3
First, a simple bar graph of the most important goal chosen is drawn, followed
by a stacked bar graph which also includes the student’s gender. We then add
a side by side bar graph that includes the student’s gender.

ggplot(StudentGoals, aes(x = Goals)) + geom_bar()

250

200

150
count

100

50

Grades Popular Sports

Goals

ggplot(StudentGoals, aes(x = Goals, fill = Gender)) + geom_bar()

96 5 Graphics in R Part 1: ggplot2

250

200

150
Gender
count

boy
girl
100

50

Grades Popular Sports

Goals

ggplot(StudentGoals, aes(x = Goals, fill = Gender)) +

geom_bar(position = "dodge")

100

Gender
count

boy
girl
50

Grades Popular Sports

Goals

In this example R counted the number of students who had each goal and
5.4 Other Types of Graphics 97

used these counts as the height of the bars. Sometimes the data contain the
bar heights as a variable. For example, we create a bar graph of India’s per
capita GDP with separate bars for each year in the data6 .

ggplot(subset(gapminder, country == "India"), aes(x = year, y = gdpPercap)) +

geom_bar(stat = "identity", color = "black", fill = "steelblue2") +
ggtitle("India's per-capita GDP")

India's per-capita GDP

2500

2000

1500
gdpPercap

1000

500

1960 1980 2000

year

5.4.4 Graphs of Functions

One way to create a plot of a mathematical function f is to create a data

frame with x values in one column and f (x) values in another column, and
then draw a line plot.

x <- seq(-pi, pi, len = 1000)

sin.data <- data.frame(x = x, y = sin(x))
ggplot(data = sin.data, aes(x = x, y = y)) + geom_line() +
scale_y_continuous(name = "sin(x)")

6R offers a large color palette, run colors() on the console to see a list of color names.
98 5 Graphics in R Part 1: ggplot2

1.0

0.5
sin(x)

0.0

-0.5

-1.0

-2 0 2
x

This method works well, but with a better understanding of functions in R we

will be able to plot mathematical functions in a simpler and more natural way.

5.5 Themes
The theme defines non-data aspects of the plot’s characteristics such as
background color, axes, and grid lines. Default themes include: theme_bw(),
theme_classic(), theme_dark(), theme_gray(), theme_light(),
theme_linedraw(), theme_minimal(), and theme_void(). Changing
the theme is as easy as adding it to your initial ggplot() call. Here I replace
the default implicit theme_bw() theme with the classic theme.

ggplot(data = sin.data, aes(x = x, y = y)) + geom_line() +

scale_y_continuous(name = "sin(x)") +
theme_classic()
5.7 Saving Graphics 99

1.0

0.5
sin(x)

0.0

-0.5

-1.0

-2 0 2
x

The ggthemes add-on package [https://github.com/jrnold/ggthemes] by

Jeffrey Arnold provides a large selection of themes beyond the eight themes
that come with ggplot2.

5.6 Saving Graphics

We often want to export our graphics to use in an external document or
share with colleagues. There are several ways to save graphics in a variety of
file formats. The ggsave() function will allow you to save your most recent
ggplot() to a variety of vector (e.g., “eps”, “ps”, “pdf”, “svg”) or raster
(e.g., “jpeg”, “tiff”, “png”, “bmp”, “wmf”) formats 7 . The subsequent call to
ggsave() saves the sin.data plot to a pdf file called “sin-plot.pdf”.

ggsave(filename = "sin-plot.pdf", device="pdf")

The ggplot() function takes additional arguments to control scale, measure-

ment units, and raster plot resolution, i.e., dots per inch (dpi).
7 Vector files comprise lines and curves known as paths, whereas raster files are comprised

of pixels. Vector images are often preferred for publication quality graphics because they
can be edited, scale well, and provide crisper detail.
100 5 Graphics in R Part 1: ggplot2

5.7 More Resources

In summary, ggplot2 provides a fairly intuitive8 framework for developing an
enormous variety of graphics. In addition to the resources mentioned at the
beginning of this chapter, there are numerous online ggplot2 resources and
galleries to get ideas for creating beautiful graphics to convey the stories in
your data. See, for example,
• https://ggplot2.tidyverse.org/
• http://www.r-graph-gallery.com/portfolio/ggplot2-package
• https://exts.ggplot2.tidyverse.org/
• http://www.cookbook-r.com/Graphs
• and of course www.google.com
While the built-in ggplot2 package documentation (accessible via the help
tab in RStudio) is helpful, the official online documentation at https://ggpl
ot2.tidyverse.org/ is particularly useful because it provides example plots
and easy navigation between related topics. The large number number of
functions and syntax in ggplot2 can be daunting. RStudio provides some
handy cheatsheets to help you along https://www.rstudio.com/resources/
cheatsheets or direct link https://www.rstudio.com/wp-content/upload
s/2016/11/ggplot2-cheatsheet-2.1.pdf.
ggplot2 also has an active mailing list at http://groups.google.com/grou
p/ggplot2. The list is an excellent resource for users at all stages of experience.
Another useful resource is stackoverflow, http://stackoverflow.com. There
is an active ggplot2 community on stackoverflow, and many common ques-
tions have already been asked and answered. When posting questions on any
programming mailing list, it is best to provide a minimal reproducible example
of your issue. The reprex https://github.com/jennybc/reprex package by
Jenny Bryan provides a convenient way to do this, and also includes advice on
creating a good example. The more information you provide about your issue,
the more likely the community is to help you.

5.7.1 Practice Problem

Check out the ggplot2 cheatsheet found at https://www.rstudio.com/wp-

content/uploads/2015/03/ggplot2-cheatsheet.pdf. Create a plot using
the iris data set using something we didn’t mention in this chapter, and
describe what is new about the plot. The plot can be anything you like, just
make sure you use something not mentioned in the chapter!
8 Like everything else in this book, it takes practice to get used to the syntax.
5.8 Exercises 101

5.8 Exercises
Exercise 6a Learning objectives: practice using ggplot2 functions; summarize
variables using graphics; introduce ggplot2 facets.
6
Working with Data

Bringing data into R, exporting data from R in a form that is readable by other
software, cleaning and reshaping data, and other data manipulation tasks are an
important and often overlooked component of data science. The book Spector
(2008), while a few years old, is still an excellent reference for data-related
issues, and is available for free at http://catalog.lib.msu.edu/record=b
7254984~S39a. And the R Data Import/Export manual, available online at
https://cran.r-project.org/doc/manuals/R-data.pdf, is an up-to-date
(and free) reference on importing a wide variety of datasets into R and on
exporting data in various forms.

6.1 Reading Data into R

Data come in a dizzying variety of forms. It might be in a proprietary format

such as an .xlsx1 Excel file, a .sav2 SPSS file, or a .mtw3 Minitab file. It might
be structured using a relational model4 comprising many tables that need
to be connected via key-value pairs. It might be a data-interchange format
such as JSON5 (JavaScript Object Notation), or a markup language such as
XML6 (Extensible Markup Language), perhaps with specialized standards for
describing ecological information, see EML7 (Ecological Metadata Language).
Both XML and EML are common data metadata formats (i.e., data that
provides information about other data). Fortunately many datasets are (or can
be) saved as plain text files, and most software can both read and write such
files, so our initial focus will be on reading plain text files into R and saving
1 https://en.wikipedia.org/wiki/Microsoft_Excel
2 https://en.wikipedia.org/wiki/SPSS
3 https://en.wikipedia.org/wiki/Minitab
4 https://en.wikipedia.org/wiki/Relational_model%7D%7Brelational%20model
5 http://www.json.org/
6 https://en.wikipedia.org/wiki/XML
7 https://en.wikipedia.org/wiki/Ecological_Metadata_Language

103
104 6 Working with Data

data from R in plain text format. RStudio provides a handy data import cheat
sheet8 for many of the read functions detailed in this section.
The foreign R package provides functions to directly read data saved in some
of the proprietary formats into R, which is sometimes unavoidable, but if
possible it is good to save data from another package as plain text and then
read this plain text file into R. In Chapter 11 methods for reading web-based
data sets into R will be discussed.
The function read.table() and its offshoots such as read.csv() are
used to read in rectangular data from a text file. For example, the file
BrainAndBody.csv contains data9 on the brain weight, body weight, and
name of some terrestrial animals. Here are the first few lines of that file:
body,brain,name
1.35,8.1,Mountain beaver
465,423,Cow
36.33,119.5,Grey wolf
27.66,115,Goat
1.04,5.5,Guinea pig
As is evident, the first line of the file contains the names of the three variables,
separated (delimited) by commas. Each subsequent line contains the body
weight, brain weight, and name of a specific terrestrial animal.
This file is accessible at the url https://www.finley-lab.com/files/data
/BrainAndBody.csv. The read.table() function is used to read these data
into an R data frame.

u.bb <- "https://www.finley-lab.com/files/data/BrainAndBody.csv"

BrainBody <- read.table(file = u.bb, header = TRUE, sep = ",",
stringsAsFactors = FALSE)
head(BrainBody)

body brain name

1 1.35 8.1 Mountain beaver
2 465.00 423.0 Cow
3 36.33 119.5 Grey wolf
4 27.66 115.0 Goat
5 1.04 5.5 Guinea pig
6 11700.00 50.0 Dipliodocus
The arguments used in this call to read.table() include:

1. file = u.bb, which tells R the loca-

8 http://www.rstudio.com/resources/cheatsheets
9 These data come from the MASS R library.
6.1 Reading Data into R 105

tion of the file. In this case the string

https://www.finley-lab.com/files/data/BrainAndBody.csv
giving the location is rather long, so it was first assigned to the
object u.bb.
2. header = TRUE, which tells R the first line of the file gives the names
of the variables.
3. sep = ",", which tells R that a comma separates the fields in the
file.
4. stringsAsFactors = FALSE which tells R not to convert character
vectors to factors.

The function read.csv() is the same as read.table() except the default

separator is a comma, whereas the default separator for read.table() is
whitespace.
The file BrainAndBody.tsv contains the same data, except a tab is used in
place of a comma to separate fields. The only change needed to read in the
data in this file is in the sep argument (and of course the file argument, since
the data are stored in a different file):

u.bb <- "https://www.finley-lab.com/files/data/BrainAndBody.tsv"

BrainBody2 <- read.table(file = u.bb, header = TRUE, sep = "\t",
stringsAsFactors = FALSE)
head(BrainBody2)

body brain name

1 1.35 8.1 Mountain beaver
2 465.00 423.0 Cow
3 36.33 119.5 Grey wolf
4 27.66 115.0 Goat
5 1.04 5.5 Guinea pig
6 11700.00 50.0 Dipliodocus
File extensions, e.g., .csv or .tsv, are naming conventions only and are there
to remind us how the columns are separated. In other words, they have no
influence on R’s file read functions.
A third file, BrainAndBody.txt, contains the same data, but also contains a
few lines of explanatory text above the names of the variables. It also uses
whitespace rather than a comma or a tab as a separator. Here are the first
several lines of the file.
This file contains data
on brain and body
weights of several terrestrial animals
106 6 Working with Data

"body" "brain" "name"

1.35 8.1 "Mountain beaver"
465 423 "Cow"
36.33 119.5 "Grey wolf"
27.66 115 "Goat"
1.04 5.5 "Guinea pig"
11700 50 "Dipliodocus"
2547 4603 "Asian elephant"
Notice that in this file the values of name are put inside of quotation marks.
This is necessary since instead R would (reasonably) assume the first line
contained the values of four variables, the values being 1.35, 8.1, Mountain,
and beaver while in reality there are only three values desired, with Mountain
beaver being the third.
To read in this file we need to tell R to skip the first five lines and to use
whitespace as the separator. The skip argument handles the first, and the sep
argument the second. First let’s see what happens if we don’t use the skip
argument.

u.bb <- "https://www.finley-lab.com/files/data/BrainAndBody.txt"

BrainBody3 <- read.table(u.bb, header = TRUE, sep = " ",
stringsAsFactors = FALSE)

Error in scan(file = file, what = what, sep = sep, quote = quote, dec = dec, : line 1 did
R assumed the first line of the file contained the variable names, since header
= TRUE was specified, and counted four including This, file, contains, and
data. So in the first line of actual data, R expected four columns containing
data plus possibly a fifth column containing row names for the data set,
and complained that “line 1 did not have 5 elements.” The error message
is somewhat mysterious, since it starts with “Error in scan.” This happens
because read.table() actually uses a more basic R function called scan()
to do the work.
Here’s how to read in the file correctly.

u.bb <- "https://www.finley-lab.com/files/data/BrainAndBody.txt"

BrainBody3 <- read.table(u.bb, header = TRUE, sep = " ",
stringsAsFactors = FALSE, skip = 4)
BrainBody3[1:10,]

body brain name

1 1.35 8.1 Mountain beaver
2 465.00 423.0 Cow
3 36.33 119.5 Grey wolf
6.2 Reading Data with Missing Observations 107

4 27.66 115.0 Goat

5 1.04 5.5 Guinea pig
6 11700.00 50.0 Dipliodocus
7 2547.00 4603.0 Asian elephant
8 187.10 419.0 Donkey
9 521.00 655.0 Horse
10 10.00 115.0 Potar monkey

6.2 Reading Data with Missing Observations

Missing data are represented in many ways. Sometimes a missing data point is
just that, i.e., the place where it should be in the file is blank. Other times
specific numbers such as −9999 or specific symbols are used. The read.table()
function has an argument na.string that allows the user to specify how missing
data is indicated in the source file.
The site http://www.wunderground.com/history/ makes weather data avail-
able for locations around the world from dates going back to 1945. The file
WeatherKLAN2014.csv contains weather data for Lansing, Michigan for the
year 2014. Here are the first few lines of that file:
EST,Max TemperatureF,Min TemperatureF, Events
1/1/14,14,9,Snow
1/2/14,13,-3,Snow
1/3/14,13,-11,Snow
1/4/14,31,13,Snow
1/5/14,29,16,Fog-Snow
1/6/14,16,-12,Fog-Snow
1/7/14,2,-13,Snow
1/8/14,17,-1,Snow
1/9/14,21,2,Snow
1/10/14,39,21,Fog-Rain-Snow
1/11/14,41,32,Fog-Rain
1/12/14,39,31,
Look at the last line, and notice that instead of an Event such as Snow or
Fog-Snow there is nothing after the comma. This observation is missing, but
rather than using an explicit code such as NA, the site just leaves that entry
blank. To read these data into R we will supply the argument na.string =
"" which tells R the file indicates missing data by leaving the appropriate
entry blank.
108 6 Working with Data

u.weather <- "https://www.finley-lab.com/files/data/WeatherKLAN2014.csv"

WeatherKLAN2014 <- read.csv(u.weather, header=TRUE,
stringsAsFactors = FALSE, na.string = "")
WeatherKLAN2014[1:15,]

EST Max.TemperatureF Min.TemperatureF

1 1/1/14 14 9
2 1/2/14 13 -3
3 1/3/14 13 -11
4 1/4/14 31 13
5 1/5/14 29 16
6 1/6/14 16 -12
7 1/7/14 2 -13
8 1/8/14 17 -1
9 1/9/14 21 2
10 1/10/14 39 21
11 1/11/14 41 32
12 1/12/14 39 31
13 1/13/14 44 34
14 1/14/14 37 26
15 1/15/14 27 18
Events
1 Snow
2 Snow
3 Snow
4 Snow
5 Fog-Snow
6 Fog-Snow
7 Snow
8 Snow
9 Snow
10 Fog-Rain-Snow
11 Fog-Rain
12 <NA>
13 Rain
14 Rain-Snow
15 Snow
6.3 Summarizing Data Frames 109

6.3 Summarizing Data Frames

Some common data tasks include variable summaries such as means or standard
deviations, transforming an existing variable, and creating new variables. As
with many tasks, there are several ways to accomplish each of these.

6.3.1 Column (and Row) Summaries

The file WeatherKLAN2014Full.csv contains a more complete set of weather

data variables than WeatherKLAN2014.csv, from the same source, http://ww
w.wunderground.com/history.

u.weather <- "https://www.finley-lab.com/files/data/WeatherKLAN2014Full.csv"

WeatherKLAN2014Full <- read.csv(u.weather, header=TRUE,
stringsAsFactors = FALSE,
na.string = "")
names(WeatherKLAN2014Full)

[1] "EST"
[2] "Max.TemperatureF"
[3] "Mean.TemperatureF"
[4] "Min.TemperatureF"
[5] "Max.Dew.PointF"
[6] "MeanDew.PointF"
[7] "Min.DewpointF"
[8] "Max.Humidity"
[9] "Mean.Humidity"
[10] "Min.Humidity"
[11] "Max.Sea.Level.PressureIn"
[12] "Mean.Sea.Level.PressureIn"
[13] "Min.Sea.Level.PressureIn"
[14] "Max.VisibilityMiles"
[15] "Mean.VisibilityMiles"
[16] "Min.VisibilityMiles"
[17] "Max.Wind.SpeedMPH"
[18] "Mean.Wind.SpeedMPH"
[19] "Max.Gust.SpeedMPH"
[20] "PrecipitationIn"
[21] "CloudCover"
[22] "Events"
[23] "WindDirDegrees"
110 6 Working with Data

How can we compute the mean for each variable? One possibility is to do this
a variable at a time:

mean(WeatherKLAN2014Full$Mean.TemperatureF)

[1] 45.78

mean(WeatherKLAN2014Full$Min.TemperatureF)

[1] 36.25

mean(WeatherKLAN2014Full$Max.TemperatureF)

[1] 54.84

##Et Cetera

This is pretty inefficient. Fortunately there is a colMeans() function which

computes the mean of each column (or a specified number of columns) in a
data frame. Some columns in the current data frame are not numeric, and
obviously we don’t want to ask R to compute means for these columns. We
use str() to investigate.

str(WeatherKLAN2014Full)

'data.frame': 365 obs. of 23 variables:

$ EST : chr "2014-1-1" "2014-1-2" "2014-1-3" "2014-1-4" ...
$ Max.TemperatureF : int 14 13 13 31 29 16 2 17 21 39 ...
$ Mean.TemperatureF : int 12 5 1 22 23 2 -5 8 12 30 ...
$ Min.TemperatureF : int 9 -3 -11 13 16 -12 -13 -1 2 21 ...
$ Max.Dew.PointF : int 9 7 2 27 27 11 -6 7 18 37 ...
$ MeanDew.PointF : int 4 4 -5 18 21 -4 -13 1 8 28 ...
$ Min.DewpointF : int 0 -8 -14 3 11 -18 -18 -6 0 19 ...
$ Max.Humidity : int 88 76 83 92 92 80 78 88 88 100 ...
$ Mean.Humidity : int 76 70 68 73 86 73 72 78 75 92 ...
$ Min.Humidity : int 63 63 53 53 80 65 65 67 62 84 ...
$ Max.Sea.Level.PressureIn : num 30.4 30.4 30.5 30.1 30 ...
$ Mean.Sea.Level.PressureIn: num 30.3 30.2 30.4 30 29.9 ...
$ Min.Sea.Level.PressureIn : num 30.2 30.1 30.1 29.9 29.7 ...
$ Max.VisibilityMiles : int 10 9 10 10 4 10 10 10 10 9 ...
$ Mean.VisibilityMiles : int 4 4 10 6 1 2 6 10 7 3 ...
$ Min.VisibilityMiles : int 1 0 5 1 0 0 1 8 2 0 ...
$ Max.Wind.SpeedMPH : int 17 22 23 28 22 31 25 15 14 17 ...
6.3 Summarizing Data Frames 111

$ Mean.Wind.SpeedMPH : int 9 13 10 15 11 18 15 7 6 10 ...

$ Max.Gust.SpeedMPH : int 22 30 32 36 30 40 31 18 17 22 ...
$ PrecipitationIn : chr "0.08" "0.01" "0.00" "0.12" ...
$ CloudCover : int 8 7 1 5 8 8 6 5 7 8 ...
$ Events : chr "Snow" "Snow" "Snow" "Snow" ...
$ WindDirDegrees : int 43 24 205 203 9 262 220 236 147 160 ...
It isn’t surprising that EST and Events are not numeric, but is surprising that
PrecipitationIn, which measures precipitation in inches, also is not numeric,
but is character. Let’s investigate further.

WeatherKLAN2014Full$PrecipitationIn[1:50]

[1] "0.08" "0.01" "0.00" "0.12" "0.78" "0.07" "T"

[8] "T" "0.01" "0.39" "0.16" "0.00" "0.00" "0.01"
[15] "T" "0.08" "T" "T" "T" "0.01" "0.00"
[22] "0.05" "T" "T" "0.07" "0.23" "0.04" "T"
[29] "T" "0.03" "T" "0.37" "T" "0.00" "T"
[36] "0.27" "0.01" "T" "0.04" "0.03" "T" "0.00"
[43] "0.00" "T" "T" "0.00" "0.02" "0.15" "0.08"
[50] "0.01"
Now it’s more clear. The original data file included T in the precipitation
column to represent a “trace” of precipitation, which is precipitation greater
than 0 but less than 0.01 inches. One possibility would be to set all these
values to "0", and then to convert the column to numeric. For now we will just
leave the PrecipitationIn column out of the columns for which we request
the mean.

colMeans(WeatherKLAN2014Full[,c(2:19, 21, 23)])

Max.TemperatureF Mean.TemperatureF
54.838 45.781
Min.TemperatureF Max.Dew.PointF
36.255 41.800
MeanDew.PointF Min.DewpointF
36.395 30.156
Max.Humidity Mean.Humidity
88.082 70.392
Min.Humidity Max.Sea.Level.PressureIn
52.200 30.130
Mean.Sea.Level.PressureIn Min.Sea.Level.PressureIn
30.015 29.904
Max.VisibilityMiles Mean.VisibilityMiles
9.896 8.249
112 6 Working with Data

Min.VisibilityMiles Max.Wind.SpeedMPH
4.825 19.101
Mean.Wind.SpeedMPH Max.Gust.SpeedMPH
8.679 NA
CloudCover WindDirDegrees
4.367 205.000

6.3.2 The apply() Function

R also has functions rowMeans(), colSums(), and rowSums(). But what if

we want to compute the median or standard deviation of columns of data,
or some other summary statistic? For this the apply() function can be used.
This function applies a user-chosen function to either the rows or columns (or
both) of a data frame. The arguments are:

1. X: the data frame of interest

2. MARGIN: specifying either rows (MARGIN = 1) or columns (MARGIN =
2)
3. FUN: the function to be applied.

apply(X = WeatherKLAN2014Full[,c(2:19, 21, 23)], MARGIN = 2, FUN = sd)

Max.TemperatureF Mean.TemperatureF
22.2130 20.9729
Min.TemperatureF Max.Dew.PointF
20.2597 19.5167
MeanDew.PointF Min.DewpointF
20.0311 20.8511
Max.Humidity Mean.Humidity
8.1910 9.3660
Min.Humidity Max.Sea.Level.PressureIn
13.9462 0.2032
Mean.Sea.Level.PressureIn Min.Sea.Level.PressureIn
0.2159 0.2360
Max.VisibilityMiles Mean.VisibilityMiles
0.5790 2.1059
Min.VisibilityMiles Max.Wind.SpeedMPH
3.8168 6.4831
Mean.Wind.SpeedMPH Max.Gust.SpeedMPH
3.8863 NA
CloudCover WindDirDegrees
2.7798 90.0673
6.4 Summarizing Data Frames 113

As with any R function the arguments don’t need to be named as long as they
are specified in the correct order, so

apply(WeatherKLAN2014Full[,c(2:19, 21, 23)], 2, sd)

has the same result.

6.3.3 Practice Problems

a. Notice the output value of NA for the column Max.Gust.SpeedMPH.

Why does this happen? Figure out a way to make the apply()
function return a numeric value for this column.
b. The apply() family of functions is extremely important in R, so it
gets two Practice Problems :) Use the apply() function to compute
the median values for all numeric columns in the iris data set.

6.3.4 Saving Typing Using with()

Consider calculating the mean of the maximum temperature values for those
days where the cloud cover is less than 4 and when the maximum humidity is
over 85. We can do this using subsetting.

mean(WeatherKLAN2014Full$Max.TemperatureF[
WeatherKLAN2014Full$CloudCover < 4 &
WeatherKLAN2014Full$Max.Humidity > 85])

[1] 69.39
While this works, it requires a lot of typing, since each time we refer to a
variable in the data set we need to preface its name by WeatherKLAN2014Full$.
The with() function tells R that we are working with a particular data frame,
and we don’t need to keep typing the name of the data frame.

with(WeatherKLAN2014Full,
mean(Max.TemperatureF[CloudCover < 4 & Max.Humidity > 85]))

[1] 69.39
114 6 Working with Data

6.4 Transforming a Data Frame

Variables are often added to, removed from, changed in, or rearranged in a
data frame. The subsetting features of R make this reasonably easy. We will
investigate this in the context of the gapminder data frame. If the gapminder
library is not yet installed, use install.packages("gapminder") to install
it locally.

library(gapminder)
str(gapminder)

tibble [1,704 x 6] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 12881816 13
$ gdpPercap: num [1:1704] 779 821 853 836 740 ...

6.4.1 Adding Variables

The data frame contains per capita GDP and population, and it might be
interesting to create a variable that gives the total GDP by multiplying these
two variables. (If we were interested in an accurate value for the total GDP we
would probably be better off getting this information directly, since it is likely
that the per capita GDP values in the data frame are rounded substantially.)

gapminder$TotalGDP <- gapminder$gdpPercap * gapminder$pop

str(gapminder)

tibble [1,704 x 7] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 12881816 13
$ gdpPercap: num [1:1704] 779 821 853 836 740 ...
$ TotalGDP : num [1:1704] 6.57e+09 7.59e+09 8.76e+09 9.65e+09 9.68e+09 ...
Analogous to the with() function, there is a function within() which can
simplify the syntax. Whereas with() does not change the data frame, within()
6.4 Transforming a Data Frame 115

can. Note, below I first remove the altered gapminder dataframe using rm()
then bring a clean copy back in by reloading the gapminder package.

rm(gapminder)
library(gapminder)
str(gapminder)

tibble [1,704 x 6] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 12881816 13
$ gdpPercap: num [1:1704] 779 821 853 836 740 ...

gapminder <- within(gapminder, TotalGDP <- gdpPercap * pop)

str(gapminder)

tibble [1,704 x 7] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 12881816 13
$ gdpPercap: num [1:1704] 779 821 853 836 740 ...
$ TotalGDP : num [1:1704] 6.57e+09 7.59e+09 8.76e+09 9.65e+09 9.68e+09 ...
A nice feature of within() is its ability to add more than one variable at
a time to a data frame. In this case the two or more formulas creating new
variables must be enclosed in braces.

gapminder <- within(gapminder, {TotalGDP <- gdpPercap * pop

lifeExpMonths <- lifeExp * 12})
str(gapminder)

tibble [1,704 x 8] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent : Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 1288181
$ gdpPercap : num [1:1704] 779 821 853 836 740 ...
$ TotalGDP : num [1:1704] 6.57e+09 7.59e+09 8.76e+09 9.65e+09 9.68e+09 ...
$ lifeExpMonths: num [1:1704] 346 364 384 408 433 ...
116 6 Working with Data

6.4.2 Removing Variables

After reflection we may realize the new variables we added to the gapminder
data frame are not useful, and should be removed.

str(gapminder)

tibble [1,704 x 8] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent : Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 1288181
$ gdpPercap : num [1:1704] 779 821 853 836 740 ...
$ TotalGDP : num [1:1704] 6.57e+09 7.59e+09 8.76e+09 9.65e+09 9.68e+09 ...
$ lifeExpMonths: num [1:1704] 346 364 384 408 433 ...

gapminder <- gapminder[1:6]

str(gapminder)

tibble [1,704 x 6] (S3: tbl_df/tbl/data.frame)

$ country : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
$ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ pop : int [1:1704] 8425333 9240934 10267083 11537966 13079460 14880372 12881816 13
$ gdpPercap: num [1:1704] 779 821 853 836 740 ...
The same result could be obtained via gapminder <- gapminder[, 1:6].
The first method uses the fact that a data frame is also a list, and uses list
subsetting methods. It is slightly preferable, since even if only one variable
is retained, the object will still be a data frame, while the other method can
return a vector in this case. Note this difference in the resulting x variable
below (again this behavior can be frustrating at times if it is not anticipated).

a <- data.frame(x = 1:3, y = c("dog", "cat", "pig"),

z = seq(from = 1, to = 2, length = 3))
a

x y z
1 1 dog 1.0
2 2 cat 1.5
3 3 pig 2.0
6.4 Transforming a Data Frame 117

a <- a[1]
a

x
1 1
2 2
3 3

a <- data.frame(x = 1:3, y = c("dog", "cat", "pig"),

z = seq(from = 1, to = 2, length = 3))
a

x y z
1 1 dog 1.0
2 2 cat 1.5
3 3 pig 2.0

a <- a[,1]
a

[1] 1 2 3
One can also use a negative sign in front of the variable number(s). For example,
a[-(2:3)] would drop the last two columns of a. Some care is needed when
removing variables using the negative sign.
An alternative approach is to set the variables you’d like to remove to NULL. For
example, a[c("y","z")] <- NULL and a[,2:3] <- NULL produce the same
result as above.

6.4.3 Practice Problem

What happens if you write a[-2:3] instead of a[-(2:3)]? Why are the
parentheses important here?

6.4.4 Transforming Variables

Consider the gapminder data again. Possibly we don’t want to add a new
variable that gives life expectancy in months, but rather want to modify the
existing variable to measure life expectancy in months. Here are two ways to
accomplish this.
118 6 Working with Data

rm(gapminder)
library(gapminder)
gapminder$lifeExp[1:5]

[1] 28.80 30.33 32.00 34.02 36.09

gapminder$lifeExp <- gapminder$lifeExp * 12

gapminder$lifeExp[1:5]

[1] 345.6 364.0 384.0 408.2 433.1

rm(gapminder)
library(gapminder)
gapminder$lifeExp[1:5]

[1] 28.80 30.33 32.00 34.02 36.09

gapminder <- within(gapminder, lifeExp <- lifeExp * 12)

gapminder$lifeExp[1:5]

[1] 345.6 364.0 384.0 408.2 433.1

6.5 Rearranging Variables

Consider the full weather data set again.

u.weather <- "https://www.finley-lab.com/files/data/WeatherKLAN2014Full.csv"

WeatherKLAN2014Full <- read.csv(u.weather, header=TRUE,
stringsAsFactors = FALSE,
na.string = "")
names(WeatherKLAN2014Full)

[1] "EST"
[2] "Max.TemperatureF"
[3] "Mean.TemperatureF"
[4] "Min.TemperatureF"
[5] "Max.Dew.PointF"
[6] "MeanDew.PointF"
6.5 Rearranging Variables 119

[7] "Min.DewpointF"
[8] "Max.Humidity"
[9] "Mean.Humidity"
[10] "Min.Humidity"
[11] "Max.Sea.Level.PressureIn"
[12] "Mean.Sea.Level.PressureIn"
[13] "Min.Sea.Level.PressureIn"
[14] "Max.VisibilityMiles"
[15] "Mean.VisibilityMiles"
[16] "Min.VisibilityMiles"
[17] "Max.Wind.SpeedMPH"
[18] "Mean.Wind.SpeedMPH"
[19] "Max.Gust.SpeedMPH"
[20] "PrecipitationIn"
[21] "CloudCover"
[22] "Events"
[23] "WindDirDegrees"
If we want the wind speed variables to come right after the date, we can again
use subsetting.

WeatherKLAN2014Full <- WeatherKLAN2014Full[c(1,17, 18, 19, 2:16, 20:23)]

names(WeatherKLAN2014Full)

[1] "EST"
[2] "Max.Wind.SpeedMPH"
[3] "Mean.Wind.SpeedMPH"
[4] "Max.Gust.SpeedMPH"
[5] "Max.TemperatureF"
[6] "Mean.TemperatureF"
[7] "Min.TemperatureF"
[8] "Max.Dew.PointF"
[9] "MeanDew.PointF"
[10] "Min.DewpointF"
[11] "Max.Humidity"
[12] "Mean.Humidity"
[13] "Min.Humidity"
[14] "Max.Sea.Level.PressureIn"
[15] "Mean.Sea.Level.PressureIn"
[16] "Min.Sea.Level.PressureIn"
[17] "Max.VisibilityMiles"
[18] "Mean.VisibilityMiles"
[19] "Min.VisibilityMiles"
[20] "PrecipitationIn"
[21] "CloudCover"
120 6 Working with Data

[22] "Events"
[23] "WindDirDegrees"

6.6 Reshaping Data

A data set can be represented in several different formats. Consider a (fictitious)

data set on incomes of three people during three different years. Here is one
representation of the data:

yearlyIncomeWide

name income1990 income2000 income2010

1 John Smith 29784 39210 41213
2 Jane Doe 56789 89321 109321
3 Albert Jones 2341 34567 56781
Here is another representation of the same data:

yearlyIncomeLong

name year income

1 John Smith income1990 29784
2 Jane Doe income1990 56789
3 Albert Jones income1990 2341
4 John Smith income2000 39210
5 Jane Doe income2000 89321
6 Albert Jones income2000 34567
7 John Smith income2010 41213
8 Jane Doe income2010 109321
9 Albert Jones income2010 56781
For hopefully obvious reasons, the first representation is called a wide repre-
sentation of the data, and the second is called a long representation. Each has
its merits. The first representation is probably easier for people to read, while
the second is often the form needed for analysis by statistical software such
as R. There are of course other representations. For example the rows and
columns could be interchanged to create a different wide representation, or
the long representation, which currently groups data by year, could group by
name instead.
Whatever the relative merits of wide and long representations of data, trans-
forming data from wide to long or long to wide is often required. As with many
6.6 Reshaping Data 121

tasks, there are several ways to accomplish this in R. We will focus on a library
called tidyr written by Hadley Wickham that performs the transformations
and more.

6.6.1 tidyr

The R library tidyr has functions for converting data between formats. To
illustrate its use, we examine a simple data set that explores the relationship
between religion and income in the United States. The data come from a Pew
survey, and are used in the tidyr documentation to illustrate transforming
data from wide to long format.

u.rel <- "https://www.finley-lab.com/files/data/religion2.csv"

religion <- read.csv(u.rel, header=TRUE, stringsAsFactors = FALSE)
head(religion)

religion under10k btw10and20k btw20and30k

1 Agnostic 27 34 60
2 Atheist 12 27 37
3 Buddhist 27 21 30
4 Catholic 418 617 732
5 DoNotKnowOrRefused 15 14 15
6 EvangelicalProt 575 869 1064
btw30and40k btw40and50k btw50and75k btw75and100k
1 81 76 137 122
2 52 35 70 73
3 34 33 58 62
4 670 638 1116 949
5 11 10 35 21
6 982 881 1486 949
btw100and150k over150k DoNotKnowOrRefused
1 109 84 96
2 59 74 76
3 39 53 54
4 792 633 1489
5 17 18 116
6 723 414 1529
As given, the columns include religion and income level, and there are counts
for each of the combinations of religion and income level. For example, there are
27 people who are Agnostic and whose income is less than 10 thousand dollars,
and there are 617 people who are Catholic and whose income is between 10
and 20 thousand dollars.
122 6 Working with Data

The pivot_longer() function can transform data from wide to long format.

library(tidyr)
religionLong <- pivot_longer(data = religion, cols = 2:11,
names_to = 'IncomeLevel', values_to = 'Frequency')
head(religionLong)

# A tibble: 6 x 3
religion IncomeLevel Frequency
<chr> <chr> <int>
1 Agnostic under10k 27
2 Agnostic btw10and20k 34
3 Agnostic btw20and30k 60
4 Agnostic btw30and40k 81
5 Agnostic btw40and50k 76
6 Agnostic btw50and75k 137

tail(religionLong)

# A tibble: 6 x 3
religion IncomeLevel Frequency
<chr> <chr> <int>
1 Unaffiliated btw40and50k 341
2 Unaffiliated btw50and75k 528
3 Unaffiliated btw75and100k 407
4 Unaffiliated btw100and150k 321
5 Unaffiliated over150k 258
6 Unaffiliated DoNotKnowOrRefused 597
To use pivot_longer() we specified the data frame (data = religion),
the columns we want to pivot into longer format (cols = 2:11), the name
we want to give the column created from the income levels (names_to =
'IncomeLevel'), and the name we want to give to the column containing the
frequency values (values_to = 'Frequency').
Columns to be pivoted into longer format can be specified by name also, and
we can also specify which columns should be omitted using a negative sign in
front of the name(s). So the following creates an equivalent data frame:

religionLong <- pivot_longer(data = religion, cols = -religion,

names_to = 'IncomeLevel', values_to = 'Frequency')
head(religionLong)

# A tibble: 6 x 3
6.6 Reshaping Data 123

religion IncomeLevel Frequency

<chr> <chr> <int>
1 Agnostic under10k 27
2 Agnostic btw10and20k 34
3 Agnostic btw20and30k 60
4 Agnostic btw30and40k 81
5 Agnostic btw40and50k 76
6 Agnostic btw50and75k 137
We now use the pivot_wider() function to widen the religionLong data set
back into the original wide format.

religionWide <- pivot_wider(data = religionLong,

names_from = IncomeLevel,
values_from = Frequency)
head(religionWide)

# A tibble: 6 x 11
religion under10k btw10and20k btw20and30k btw30and40k
<chr> <int> <int> <int> <int>
1 Agnostic 27 34 60 81
2 Atheist 12 27 37 52
3 Buddhist 27 21 30 34
4 Catholic 418 617 732 670
5 DoNotKn~ 15 14 15 11
6 Evangel~ 575 869 1064 982
# ... with 6 more variables: btw40and50k <int>,
# btw50and75k <int>, btw75and100k <int>,
# btw100and150k <int>, over150k <int>,
# DoNotKnowOrRefused <int>
Here we specify the data frame (data = religionLong), the column
(names_from = IncomeLevel) to get the name of the output column, and
the column of values (values_from = Frequency) to get the cell values from.
As can be seen, this particular call to pivot_longer() yields the original data
frame.
tidyr provides two other useful functions to separate and unite variables based
on some deliminator. Consider again the yearlyIncomeWide table. Say we
want to split the name variable into first and last name. This can be done using
the separate() function.

firstLast <- separate(data = yearlyIncomeLong, col = name,

into = c("first", "last"), sep="\\s")
print(firstLast)
124 6 Working with Data

first last year income

1 John Smith income1990 29784
2 Jane Doe income1990 56789
3 Albert Jones income1990 2341
4 John Smith income2000 39210
5 Jane Doe income2000 89321
6 Albert Jones income2000 34567
7 John Smith income2010 41213
8 Jane Doe income2010 109321
9 Albert Jones income2010 56781
Now say, you’re not happy with that and you want to combine the name
column again, but this time separate the first and last name with a underscore.
This is done using the unite() function.

unite(firstLast, col=name, first, last, sep="_")

name year income

1 John_Smith income1990 29784
2 Jane_Doe income1990 56789
3 Albert_Jones income1990 2341
4 John_Smith income2000 39210
5 Jane_Doe income2000 89321
6 Albert_Jones income2000 34567
7 John_Smith income2010 41213
8 Jane_Doe income2010 109321
9 Albert_Jones income2010 56781

6.6.2 Practice Problem

The separate() function from the tidyr library is especially useful when
working with real data as multiple pieces of information can be combined into
one column in a data set. For example, consider the following data frame

data.birds

birds recordingInfo
1 10 LA-2017-01-01
2 38 DF-2011-03-02
3 29 OG-2078-05-11
4 88 YA-2000-11-18
5 42 LA-2019-03-17
6 177 OG-2016-10-10
7 200 YA-2001-03-22
6.7 Manipulating Data with dplyr 125

“RecordingInfo” that contains the site where data was collected, as well as the
year, month, and date the data were recorded. The data are coded as follows:
site-year-month-day. Write a line of code that will extract the desired data
from the data.birds data frame into separate columns named site, year,
month and day.

6.7 Manipulating Data with dplyr

Much of the effort (a figure of 80% is sometimes suggested) in data analysis
is spent cleaning the data and getting it ready for analysis. Having effective
tools for this task can save substantial time and effort. The R package dplyr
written by Hadley Wickham is designed, in Hadley’s words, to be “a grammar
of data manipulation, providing a consistent set of verbs that help you solve
the most common data manipulation challenges.” Casting data analysis tasks
in terms of “grammar” should be familiar from our work with the ggplot2
package, which was also authored by Hadley. Functions provided by dplyr do
in fact capture key data analysis actions (i.e., verbs). These functions include
• mutate() adds new variables that are functions of existing variables
• select() picks variables based on their names
• filter() picks cases based on their values
• summarize() reduces multiple values down to a single summary
• arrange() changes the ordering of the rows.
These all combine naturally with a group_by() function that allows you to
perform any operation grouped by values of one or more variables. All the
tasks done using dplyr can be accomplished using tools already covered in
this text; however, dplyr’s functions provide a potentially more efficient and
convenient framework to accomplish these tasks. RStudio provides a convenient
data wrangling cheat sheet10 that covers many aspects of the tidyr and dplyr
packages.
This somewhat long section on dplyr adapts the nice introduction by Jenny
Bryan, available at https://stat545.com/dplyr-intro.html.
10 https://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-

cheatsheet.pdf
126 6 Working with Data

6.7.1 Improved Data Frames

The dplyr package provides a couple functions that offer improvements on

data frames. First, tibble creates a tibble from a series of vectors11 . A tibble
has two advantages over a data frame. First, when printing, it only prints
the first ten rows and the columns that fit on the page, as well as some
additional information about the table’s dimension, data type of variables,
and non-printed columns. Second, recall that subsetting a data frame can
sometimes return a vector rather than a data frame (if only one row or column
is the result of the subset), a tibble does not have this behavior. Second, we
can use the as_tibble() function to convert an existing data frame or list
into a tibble. Here is an example using the mtcars data frame.13

library(dplyr)
head(mtcars)

mpg cyl disp hp drat wt qsec

Mazda RX4 21.0 6 160 110 3.90 2.620 16.46
Mazda RX4 Wag 21.0 6 160 110 3.90 2.875 17.02
Datsun 710 22.8 4 108 93 3.85 2.320 18.61
Hornet 4 Drive 21.4 6 258 110 3.08 3.215 19.44
Hornet Sportabout 18.7 8 360 175 3.15 3.440 17.02
Valiant 18.1 6 225 105 2.76 3.460 20.22
vs am gear carb
Mazda RX4 0 1 4 4
Mazda RX4 Wag 0 1 4 4
Datsun 710 1 1 4 1
Hornet 4 Drive 1 0 3 1
Hornet Sportabout 0 0 3 2
Valiant 1 0 3 1

mtcars[,1]

[1] 21.0 21.0 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2
[11] 17.8 16.4 17.3 15.2 10.4 10.4 14.7 32.4 30.4 33.9
[21] 21.5 15.5 15.2 13.3 19.2 27.3 26.0 30.4 15.8 19.7
[31] 15.0 21.4

11 Reminds me of The Trouble with Tribbles12

13 The text printed immediately after library(dplyr) means the stats and base packages,
which are automatically loaded when you start R, have functions with the same name as
functions in dplyr. So, for example, if you call the filter() or lag() functions, R will
use library(dplyr)‘s functions. Use the :: operator to explicity identify which packages’
function you want to use, e.g., if you want stats’s lag() then call stats::lag().
6.7 Manipulating Data with dplyr 127

mtcarsTbl <- as_tibble(mtcars)

head(mtcarsTbl)

# A tibble: 6 x 11
mpg cyl disp hp drat wt qsec vs am
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 21 6 160 110 3.9 2.62 16.5 0 1
2 21 6 160 110 3.9 2.88 17.0 0 1
3 22.8 4 108 93 3.85 2.32 18.6 1 1
4 21.4 6 258 110 3.08 3.22 19.4 1 0
5 18.7 8 360 175 3.15 3.44 17.0 0 0
6 18.1 6 225 105 2.76 3.46 20.2 1 0
# ... with 2 more variables: gear <dbl>, carb <dbl>

mtcarsTbl[,1]

# A tibble: 32 x 1
mpg
<dbl>
1 21
2 21
3 22.8
4 21.4
5 18.7
6 18.1
7 14.3
8 24.4
9 22.8
10 19.2
# ... with 22 more rows
As seen above, note that once the data frame is reduced to one dimension by
subsetting to one column, it is no longer a data frame and has been simplified to
a vector. This might not seem like a big deal; however, it can be very frustrating
and potentially break your code when you expect an object to behave like a
data frame and it doesn’t because it’s now a vector. Alternatively, once we
convert religionWide to a tibble via the as_tibble() function the object
remains a data frame even when subsetting down to one dimension (there is
no automatic simplification). Converting data frames using as_tibble() is
not required for using dplyr but is convenient. Also, it is important to note
that tibble is simply a wrapper around a data frame that provides some
additional behaviors. The newly formed tibble object will still behave like
a data frame (because it technically still is a data frame) but will have some
added niceties (some of which are illustrated below).
128 6 Working with Data

6.7.2 Filtering Data by Row

Recall the gapminder data. These data are available in tab-separated format
in gapminder.tsv, and can be read in using read.delim() (or the related
read functions described previously). The read.delim() function defaults to
header = TRUE so this doesn’t need to be specified explicitly. In this section
we will be working with the gapminder data often, so we will use a short name
for the data frame to save typing.

u.gm <- "https://www.finley-lab.com/files/data/gapminder.tsv"

gm <- read.delim(u.gm)
gm <- as_tibble(gm)
str(gm)

tibble [1,704 x 6] (S3: tbl_df/tbl/data.frame)

$ country : chr [1:1704] "Afghanistan" "Afghanistan" "Afghanistan" "Afghanistan" ...
$ year : int [1:1704] 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
$ pop : num [1:1704] 8425333 9240934 10267083 11537966 13079460 ...
$ continent: chr [1:1704] "Asia" "Asia" "Asia" "Asia" ...
$ lifeExp : num [1:1704] 28.8 30.3 32 34 36.1 ...
$ gdpPercap: num [1:1704] 779 821 853 836 740 ...

head(gm)

# A tibble: 6 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Afghanistan 1952 8.43e6 Asia 28.8 779.
2 Afghanistan 1957 9.24e6 Asia 30.3 821.
3 Afghanistan 1962 1.03e7 Asia 32.0 853.
4 Afghanistan 1967 1.15e7 Asia 34.0 836.
5 Afghanistan 1972 1.31e7 Asia 36.1 740.
6 Afghanistan 1977 1.49e7 Asia 38.4 786.
Filtering helps us to examine subsets of the data such as data from a particular
country, from several specified countries, from certain years, from countries
with certain populations, etc. Some examples:

filter(gm, country == "Brazil")

# A tibble: 12 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Brazil 1952 56602560 Americas 50.9 2109.
6.7 Manipulating Data with dplyr 129

2 Brazil 1957 65551171 Americas 53.3 2487.

3 Brazil 1962 76039390 Americas 55.7 3337.
4 Brazil 1967 88049823 Americas 57.6 3430.
5 Brazil 1972 100840058 Americas 59.5 4986.
6 Brazil 1977 114313951 Americas 61.5 6660.
7 Brazil 1982 128962939 Americas 63.3 7031.
8 Brazil 1987 142938076 Americas 65.2 7807.
9 Brazil 1992 155975974 Americas 67.1 6950.
10 Brazil 1997 168546719 Americas 69.4 7958.
11 Brazil 2002 179914212 Americas 71.0 8131.
12 Brazil 2007 190010647 Americas 72.4 9066.

filter(gm, country %in% c("Brazil", "Mexico"))

# A tibble: 24 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Brazil 1952 56602560 Americas 50.9 2109.
2 Brazil 1957 65551171 Americas 53.3 2487.
3 Brazil 1962 76039390 Americas 55.7 3337.
4 Brazil 1967 88049823 Americas 57.6 3430.
5 Brazil 1972 100840058 Americas 59.5 4986.
6 Brazil 1977 114313951 Americas 61.5 6660.
7 Brazil 1982 128962939 Americas 63.3 7031.
8 Brazil 1987 142938076 Americas 65.2 7807.
9 Brazil 1992 155975974 Americas 67.1 6950.
10 Brazil 1997 168546719 Americas 69.4 7958.
# ... with 14 more rows

filter(gm, country %in% c("Brazil", "Mexico") & year %in% c(1952, 1972))

# A tibble: 4 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Brazil 1952 56602560 Americas 50.9 2109.
2 Brazil 1972 100840058 Americas 59.5 4986.
3 Mexico 1952 30144317 Americas 50.8 3478.
4 Mexico 1972 55984294 Americas 62.4 6809.

filter(gm, pop > 300000000)

# A tibble: 25 x 6
country year pop continent lifeExp gdpPercap
130 6 Working with Data

<chr> <int> <dbl> <chr> <dbl> <dbl>

1 China 1952 5.56e8 Asia 44 400.
2 China 1957 6.37e8 Asia 50.5 576.
3 China 1962 6.66e8 Asia 44.5 488.
4 China 1967 7.55e8 Asia 58.4 613.
5 China 1972 8.62e8 Asia 63.1 677.
6 China 1977 9.43e8 Asia 64.0 741.
7 China 1982 1.00e9 Asia 65.5 962.
8 China 1987 1.08e9 Asia 67.3 1379.
9 China 1992 1.16e9 Asia 68.7 1656.
10 China 1997 1.23e9 Asia 70.4 2289.
# ... with 15 more rows

filter(gm, pop > 300000000 & year == 2007)

# A tibble: 3 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 China 2007 1.32e9 Asia 73.0 4959.
2 India 2007 1.11e9 Asia 64.7 2452.
3 United St~ 2007 3.01e8 Americas 78.2 42952.
Notice the full results are not printed. For example, when we asked for the
data for Brazil and Mexico, only the first ten rows were printed. This is an
effect of using the as_tibble() function. Of course if we wanted to analyze
the results (as we will below) the full set of data would be available.

6.7.3 Selecting variables by column

Continuing with the gapminder data, another common task is to restrict

attention to some subset of variables in the data set. The select() function
does this.

select(gm, country, year, lifeExp)

# A tibble: 1,704 x 3
country year lifeExp
<chr> <int> <dbl>
1 Afghanistan 1952 28.8
2 Afghanistan 1957 30.3
3 Afghanistan 1962 32.0
4 Afghanistan 1967 34.0
5 Afghanistan 1972 36.1
6.7 Manipulating Data with dplyr 131

6 Afghanistan 1977 38.4

7 Afghanistan 1982 39.9
8 Afghanistan 1987 40.8
9 Afghanistan 1992 41.7
10 Afghanistan 1997 41.8
# ... with 1,694 more rows

select(gm, 2:4)

# A tibble: 1,704 x 3
year pop continent
<int> <dbl> <chr>
1 1952 8425333 Asia
2 1957 9240934 Asia
3 1962 10267083 Asia
4 1967 11537966 Asia
5 1972 13079460 Asia
6 1977 14880372 Asia
7 1982 12881816 Asia
8 1987 13867957 Asia
9 1992 16317921 Asia
10 1997 22227415 Asia
# ... with 1,694 more rows

select(gm, -c(2,3,4))

# A tibble: 1,704 x 3
country lifeExp gdpPercap
<chr> <dbl> <dbl>
1 Afghanistan 28.8 779.
2 Afghanistan 30.3 821.
3 Afghanistan 32.0 853.
4 Afghanistan 34.0 836.
5 Afghanistan 36.1 740.
6 Afghanistan 38.4 786.
7 Afghanistan 39.9 978.
8 Afghanistan 40.8 852.
9 Afghanistan 41.7 649.
10 Afghanistan 41.8 635.
# ... with 1,694 more rows

select(gm, starts_with("c"))

# A tibble: 1,704 x 2
132 6 Working with Data

country continent
<chr> <chr>
1 Afghanistan Asia
2 Afghanistan Asia
3 Afghanistan Asia
4 Afghanistan Asia
5 Afghanistan Asia
6 Afghanistan Asia
7 Afghanistan Asia
8 Afghanistan Asia
9 Afghanistan Asia
10 Afghanistan Asia
# ... with 1,694 more rows
Notice a few things. Variables can be selected by name or column number. As
usual, a negative sign tells R to leave something out. And there are special
functions such as starts_with that provide ways to match part of a variable’s
name.

6.7.4 Practice Problem

Use the contains() function to select only the columns that contain a c in
the gapminder data set.

6.7.5 Pipes

Consider selecting the country, year, and population for countries in Asia
or Europe. One possibility is to nest a filter() function inside a select()
function.

select(filter(gm, continent %in% c("Asia", "Europe")), country, year, pop)

# A tibble: 756 x 3
country year pop
<chr> <int> <dbl>
1 Afghanistan 1952 8425333
2 Afghanistan 1957 9240934
3 Afghanistan 1962 10267083
4 Afghanistan 1967 11537966
5 Afghanistan 1972 13079460
6 Afghanistan 1977 14880372
7 Afghanistan 1982 12881816
8 Afghanistan 1987 13867957
6.7 Manipulating Data with dplyr 133

9 Afghanistan 1992 16317921

10 Afghanistan 1997 22227415
# ... with 746 more rows
Even a two-step process like this becomes hard to follow in this nested form,
and often we will want to perform more than two operations. There is a nice
feature in dplyr that allows us to “feed” results of one function into the first
argument of a subsequent function. Another way of saying this is that we are
“piping” the results into another function. The %>% operator does the piping.
Here we again restrict attention to country, year, and population for countries
in Asia or Europe14 .

gm %>%
filter(continent %in% c("Asia", "Europe")) %>%
select(country, year, pop)

# A tibble: 756 x 3
country year pop
<chr> <int> <dbl>
1 Afghanistan 1952 8425333
2 Afghanistan 1957 9240934
3 Afghanistan 1962 10267083
4 Afghanistan 1967 11537966
5 Afghanistan 1972 13079460
6 Afghanistan 1977 14880372
7 Afghanistan 1982 12881816
8 Afghanistan 1987 13867957
9 Afghanistan 1992 16317921
10 Afghanistan 1997 22227415
# ... with 746 more rows
It can help to think of %>% as representing the word “then”. The above can
be read as, “Start with the data frame gm then filter it to select data from
the continents Asia and Europe then select the variables country, year, and
population from these data”.
The pipe operator %>% is not restricted to functions in dplyr. In fact the
pipe operator itself was introduced in another package called magrittr, but is
included in dplyr as a convenience.
14 Notice the indentation used in the code. This is not necessary, as the code could be all

on one line, but I often find it easier to read in this more organized format
134 6 Working with Data

6.7.6 Arranging Data by Row

By default the gapminder data are arranged by country and then by year.

head(gm, 15)

# A tibble: 15 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Afghanist~ 1952 8.43e6 Asia 28.8 779.
2 Afghanist~ 1957 9.24e6 Asia 30.3 821.
3 Afghanist~ 1962 1.03e7 Asia 32.0 853.
4 Afghanist~ 1967 1.15e7 Asia 34.0 836.
5 Afghanist~ 1972 1.31e7 Asia 36.1 740.
6 Afghanist~ 1977 1.49e7 Asia 38.4 786.
7 Afghanist~ 1982 1.29e7 Asia 39.9 978.
8 Afghanist~ 1987 1.39e7 Asia 40.8 852.
9 Afghanist~ 1992 1.63e7 Asia 41.7 649.
10 Afghanist~ 1997 2.22e7 Asia 41.8 635.
11 Afghanist~ 2002 2.53e7 Asia 42.1 727.
12 Afghanist~ 2007 3.19e7 Asia 43.8 975.
13 Albania 1952 1.28e6 Europe 55.2 1601.
14 Albania 1957 1.48e6 Europe 59.3 1942.
15 Albania 1962 1.73e6 Europe 64.8 2313.
Possibly arranging the data by year and then country would be desired. The
arrange() function makes this easy. We will again use pipes.

gm %>% arrange(year, country)

# A tibble: 1,704 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Afghanist~ 1952 8.43e6 Asia 28.8 779.
2 Albania 1952 1.28e6 Europe 55.2 1601.
3 Algeria 1952 9.28e6 Africa 43.1 2449.
4 Angola 1952 4.23e6 Africa 30.0 3521.
5 Argentina 1952 1.79e7 Americas 62.5 5911.
6 Australia 1952 8.69e6 Oceania 69.1 10040.
7 Austria 1952 6.93e6 Europe 66.8 6137.
8 Bahrain 1952 1.20e5 Asia 50.9 9867.
9 Bangladesh 1952 4.69e7 Asia 37.5 684.
10 Belgium 1952 8.73e6 Europe 68 8343.
# ... with 1,694 more rows
6.7 Manipulating Data with dplyr 135

How about the data for Rwanda, arranged in order of life expectancy.

gm %>%
filter(country == "Rwanda") %>%
arrange(lifeExp)

# A tibble: 12 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Rwanda 1992 7290203 Africa 23.6 737.
2 Rwanda 1997 7212583 Africa 36.1 590.
3 Rwanda 1952 2534927 Africa 40 493.
4 Rwanda 1957 2822082 Africa 41.5 540.
5 Rwanda 1962 3051242 Africa 43 597.
6 Rwanda 2002 7852401 Africa 43.4 786.
7 Rwanda 1987 6349365 Africa 44.0 848.
8 Rwanda 1967 3451079 Africa 44.1 511.
9 Rwanda 1972 3992121 Africa 44.6 591.
10 Rwanda 1977 4657072 Africa 45 670.
11 Rwanda 1982 5507565 Africa 46.2 882.
12 Rwanda 2007 8860588 Africa 46.2 863.
Possibly we want these data to be in decreasing (descending) order. Here,
desc() is one of many dplyr helper functions.

gm %>%
filter(country == "Rwanda") %>%
arrange(desc(lifeExp))

# A tibble: 12 x 6
country year pop continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Rwanda 2007 8860588 Africa 46.2 863.
2 Rwanda 1982 5507565 Africa 46.2 882.
3 Rwanda 1977 4657072 Africa 45 670.
4 Rwanda 1972 3992121 Africa 44.6 591.
5 Rwanda 1967 3451079 Africa 44.1 511.
6 Rwanda 1987 6349365 Africa 44.0 848.
7 Rwanda 2002 7852401 Africa 43.4 786.
8 Rwanda 1962 3051242 Africa 43 597.
9 Rwanda 1957 2822082 Africa 41.5 540.
10 Rwanda 1952 2534927 Africa 40 493.
11 Rwanda 1997 7212583 Africa 36.1 590.
12 Rwanda 1992 7290203 Africa 23.6 737.
136 6 Working with Data

Possibly we want to include only the year and life expectancy, to make the
message more stark.

gm %>%
filter(country == "Rwanda") %>%
select(year, lifeExp) %>%
arrange(desc(lifeExp))

# A tibble: 12 x 2
year lifeExp
<int> <dbl>
1 2007 46.2
2 1982 46.2
3 1977 45
4 1972 44.6
5 1967 44.1
6 1987 44.0
7 2002 43.4
8 1962 43
9 1957 41.5
10 1952 40
11 1997 36.1
12 1992 23.6
For analyzing data in R, the order shouldn’t matter. But for presentation to
human eyes, the order is important.

6.7.7 Practice Problem

It is worth your while to get comfortable with using pipes. Here is some
hard-to-read code. Convert it into more easier to read code by using pipes.

arrange(select(filter(gm, country == "Afghanistan"),

c("year", "lifeExp")), desc(lifeExp))

# A tibble: 12 x 2
year lifeExp
<int> <dbl>
1 2007 43.8
2 2002 42.1
3 1997 41.8
4 1992 41.7
5 1987 40.8
6.7 Manipulating Data with dplyr 137

6 1982 39.9
7 1977 38.4
8 1972 36.1
9 1967 34.0
10 1962 32.0
11 1957 30.3
12 1952 28.8

6.7.8 Renaming Variables

The dplyr package has a rename function that makes renaming variables in a
data frame quite easy.

gm <- rename(gm, population = pop)

head(gm)

# A tibble: 6 x 6
country year population continent lifeExp gdpPercap
<chr> <int> <dbl> <chr> <dbl> <dbl>
1 Afghani~ 1952 8425333 Asia 28.8 779.
2 Afghani~ 1957 9240934 Asia 30.3 821.
3 Afghani~ 1962 10267083 Asia 32.0 853.
4 Afghani~ 1967 11537966 Asia 34.0 836.
5 Afghani~ 1972 13079460 Asia 36.1 740.
6 Afghani~ 1977 14880372 Asia 38.4 786.

6.7.9 Data Summaries and Grouping

The summarize() function computes summary statistics using user provided

functions for one or more columns of data in a data frame.

summarize(gm, meanpop = mean(population), medpop = median(population))

# A tibble: 1 x 2
meanpop medpop
<dbl> <dbl>
1 29601212. 7023596.

##or
gm %>%
summarize(meanpop = mean(population), medpop = median(population))
138 6 Working with Data

# A tibble: 1 x 2
meanpop medpop
<dbl> <dbl>
1 29601212. 7023596.
Often we want summaries for specific components of the data. For example,
we might want the median life expectancy for each continent separately. One
option is subsetting:

median(gm$lifeExp[gm$continent == "Africa"])

[1] 47.79

median(gm$lifeExp[gm$continent == "Asia"])

[1] 61.79

median(gm$lifeExp[gm$continent == "Europe"])

[1] 72.24

median(gm$lifeExp[gm$continent == "Americas"])

[1] 67.05

median(gm$lifeExp[gm$continent == "Oceania"])

[1] 73.66
The group_by() function makes this easier, and makes the output more useful.

gm %>%
group_by(continent) %>%
summarize(medLifeExp = median(lifeExp))

`summarise()` ungrouping output (override with `.groups` argument)

# A tibble: 5 x 2
continent medLifeExp
<chr> <dbl>
1 Africa 47.8
2 Americas 67.0
3 Asia 61.8
6.7 Manipulating Data with dplyr 139

4 Europe 72.2
5 Oceania 73.7
Or if we want the results ordered by the median life expectancy:

gm %>%
group_by(continent) %>%
summarize(medLifeExp = median(lifeExp)) %>%
arrange(medLifeExp)

`summarise()` ungrouping output (override with `.groups` argument)

# A tibble: 5 x 2
continent medLifeExp
<chr> <dbl>
1 Africa 47.8
2 Asia 61.8
3 Americas 67.0
4 Europe 72.2
5 Oceania 73.7
As another example, we calculate the number of observations we have per
continent (using the n() helper function), and then, among continents, how
many distinct countries are represented (using n_distinct()).

gm %>%
group_by(continent) %>%
summarize(numObs = n())

`summarise()` ungrouping output (override with `.groups` argument)

# A tibble: 5 x 2
continent numObs
<chr> <int>
1 Africa 624
2 Americas 300
3 Asia 396
4 Europe 360
5 Oceania 24

gm %>%
group_by(continent) %>%
summarize(n_obs = n(), n_countries = n_distinct(country))

`summarise()` ungrouping output (override with `.groups` argument)

140 6 Working with Data

# A tibble: 5 x 3
continent n_obs n_countries
<chr> <int> <int>
1 Africa 624 52
2 Americas 300 25
3 Asia 396 33
4 Europe 360 30
5 Oceania 24 2
Here is a bit more involved example that calculates the minimum and maximum
life expectancies for countries in Africa by year.

gm %>%
filter(continent == "Africa") %>%
group_by(year) %>%
summarize(min_lifeExp = min(lifeExp), max_lifeExp = max(lifeExp))

`summarise()` ungrouping output (override with `.groups` argument)

# A tibble: 12 x 3
year min_lifeExp max_lifeExp
<int> <dbl> <dbl>
1 1952 30 52.7
2 1957 31.6 58.1
3 1962 32.8 60.2
4 1967 34.1 61.6
5 1972 35.4 64.3
6 1977 36.8 67.1
7 1982 38.4 69.9
8 1987 39.9 71.9
9 1992 23.6 73.6
10 1997 36.1 74.8
11 2002 39.2 75.7
12 2007 39.6 76.4
This is interesting, but the results don’t include the countries that achieved the
minimum and maximum life expectancies. Here is one way to achieve that. We
will start with the minimum life expectancy. Note the rank of the minimum
value will be 1.

gm %>%
select(country, continent, year, lifeExp) %>%
group_by(year) %>%
arrange(year) %>%
filter(rank(lifeExp) == 1)
6.7 Manipulating Data with dplyr 141

# A tibble: 12 x 4
# Groups: year [12]
country continent year lifeExp
<chr> <chr> <int> <dbl>
1 Afghanistan Asia 1952 28.8
2 Afghanistan Asia 1957 30.3
3 Afghanistan Asia 1962 32.0
4 Afghanistan Asia 1967 34.0
5 Sierra Leone Africa 1972 35.4
6 Cambodia Asia 1977 31.2
7 Sierra Leone Africa 1982 38.4
8 Angola Africa 1987 39.9
9 Rwanda Africa 1992 23.6
10 Rwanda Africa 1997 36.1
11 Zambia Africa 2002 39.2
12 Swaziland Africa 2007 39.6
Next we add the maximum life expectancy. Here we need to better understand
the desc() function, which will transform a vector into a numeric vector which
will be sorted in descending order. Here are some examples.

desc(1:5)

[1] -1 -2 -3 -4 -5

desc(c(2,3,1,5,6,-4))

[1] -2 -3 -1 -5 -6 4

desc(c("a", "c", "b", "w", "e"))

[1] -1 -3 -2 -5 -4
We now use this to extract the maximum life expectancy. Recall that |
represents “or”. Also by default only the first few rows of a tibble object will
be printed. To see all the rows we pipe the output to print(n = 24) to ask
for all 24 rows to be printed.

gm %>%
select(country, continent, year, lifeExp) %>%
group_by(year) %>%
arrange(year) %>%
filter(rank(lifeExp) == 1 | rank(desc(lifeExp)) == 1) %>%
print(n=24)
142 6 Working with Data

# A tibble: 24 x 4
# Groups: year [12]
country continent year lifeExp
<chr> <chr> <int> <dbl>
1 Afghanistan Asia 1952 28.8
2 Norway Europe 1952 72.7
3 Afghanistan Asia 1957 30.3
4 Iceland Europe 1957 73.5
5 Afghanistan Asia 1962 32.0
6 Iceland Europe 1962 73.7
7 Afghanistan Asia 1967 34.0
8 Sweden Europe 1967 74.2
9 Sierra Leone Africa 1972 35.4
10 Sweden Europe 1972 74.7
11 Cambodia Asia 1977 31.2
12 Iceland Europe 1977 76.1
13 Japan Asia 1982 77.1
14 Sierra Leone Africa 1982 38.4
15 Angola Africa 1987 39.9
16 Japan Asia 1987 78.7
17 Japan Asia 1992 79.4
18 Rwanda Africa 1992 23.6
19 Japan Asia 1997 80.7
20 Rwanda Africa 1997 36.1
21 Japan Asia 2002 82
22 Zambia Africa 2002 39.2
23 Japan Asia 2007 82.6
24 Swaziland Africa 2007 39.6

6.7.10 Creating New Variables

The $ notation provides a simple way to create new variables in a data frame.
The mutate() function provides another, sometimes cleaner way to do this.
We will use mutate() along with the lag() function to investigate changes in
life expectancy over five years for the gapminder data. We’ll do this in a few
steps. First, we create a variable that measures the change in life expectancy
and remove the population and GDP variables that are not of interest. We
have to be careful to first group by country, since we want to calculate the
change in life expectancy by country.

gm %>%
group_by(country) %>%
6.7 Manipulating Data with dplyr 143

mutate(changeLifeExp = lifeExp - lag(lifeExp, order_by = year)) %>%

select(-c(population, gdpPercap))

# A tibble: 1,704 x 5
# Groups: country [142]
country year continent lifeExp changeLifeExp
<chr> <int> <chr> <dbl> <dbl>
1 Afghanistan 1952 Asia 28.8 NA
2 Afghanistan 1957 Asia 30.3 1.53
3 Afghanistan 1962 Asia 32.0 1.66
4 Afghanistan 1967 Asia 34.0 2.02
5 Afghanistan 1972 Asia 36.1 2.07
6 Afghanistan 1977 Asia 38.4 2.35
7 Afghanistan 1982 Asia 39.9 1.42
8 Afghanistan 1987 Asia 40.8 0.968
9 Afghanistan 1992 Asia 41.7 0.852
10 Afghanistan 1997 Asia 41.8 0.0890
# ... with 1,694 more rows
Next, summarize by computing the largest drop in life expectancy.

gm %>%
group_by(country) %>%
mutate(changeLifeExp = lifeExp - lag(lifeExp, order_by = year)) %>%
select(-c(population, gdpPercap)) %>%
summarize(largestDropLifeExp = min(changeLifeExp))

`summarise()` ungrouping output (override with `.groups` argument)

# A tibble: 142 x 2
country largestDropLifeExp
<chr> <dbl>
1 Afghanistan NA
2 Albania NA
3 Algeria NA
4 Angola NA
5 Argentina NA
6 Australia NA
7 Austria NA
8 Bahrain NA
9 Bangladesh NA
10 Belgium NA
# ... with 132 more rows
144 6 Working with Data

Oops. We forgot that since we don’t have data from before 1952, the first drop
will be NA. Let’s try again.

gm %>%
group_by(country) %>%
mutate(changeLifeExp = lifeExp - lag(lifeExp, order_by = year)) %>%
select(-c(population, gdpPercap)) %>%
summarize(largestDropLifeExp = min(changeLifeExp, na.rm = TRUE))

`summarise()` ungrouping output (override with `.groups` argument)

# A tibble: 142 x 2
country largestDropLifeExp
<chr> <dbl>
1 Afghanistan 0.0890
2 Albania -0.419
3 Algeria 1.31
4 Angola -0.036
5 Argentina 0.492
6 Australia 0.170
7 Austria 0.490
8 Bahrain 0.84
9 Bangladesh 1.67
10 Belgium 0.5
# ... with 132 more rows
That’s not quite what we wanted. We could arrange the results by the life
expectancy drop, but it would be good to have both the continent and year
printed out also. So we’ll take a slightly different approach, by arranging the
results in increasing order.

gm %>%
group_by(country) %>%
mutate(changeLifeExp = lifeExp - lag(lifeExp, order_by = year)) %>%
select(-c(population, gdpPercap)) %>%
arrange(changeLifeExp)

# A tibble: 1,704 x 5
# Groups: country [142]
country year continent lifeExp changeLifeExp
<chr> <int> <chr> <dbl> <dbl>
1 Rwanda 1992 Africa 23.6 -20.4
2 Zimbabwe 1997 Africa 46.8 -13.6
3 Lesotho 2002 Africa 44.6 -11.0
4 Swaziland 2002 Africa 43.9 -10.4
6.8 Exercises 145

5 Botswana 1997 Africa 52.6 -10.2

6 Cambodia 1977 Asia 31.2 -9.10
7 Namibia 2002 Africa 51.5 -7.43
8 South Africa 2002 Africa 53.4 -6.87
9 Zimbabwe 2002 Africa 40.0 -6.82
10 China 1962 Asia 44.5 -6.05
# ... with 1,694 more rows

6.7.11 Practice Problem

As you progress in your data science related career, we are sure you will find
dplyr as one of the most useful packages for your initial data exploration.
Here is one more Practice Problem to get more comfortable with the syntax.
Recall the iris data set we have worked with multiple times. We want to look
at the ratio of Sepal.Length to Petal.Length in the three different species.
Write a series of dplyr statements that groups the data by species, creates a new
column called s.p.ratio that is the Sepal.Length divided by Petal.Length,
then computes the mean of this column for each species in a column called
mean.ratio. Display the data in descending order of mean.ratio.

6.8 Exercises

Exercise 7 Learning objectives: introduce with(), tapply(), and cut()

functions; summarize data using the table() function with logical subsetting;
practice using factor data types.
Exercise 8 Learning objectives: work with messy data; import data from an
external spreadsheet; practice using functions in tidyr and graphing func-
tions.}
Exercise 9 Learning objectives: work with several key dplyr functions; ma-
nipulate data frames (actually tibbles); summarize and visualize data from
large data files.
7
Functions and Programming

We have been working with a wide variety of R functions, from simple functions
such as mean() and sd() to more complex functions such as ggplot() and
apply(). Gaining a better understanding of existing functions and the ability
to write your own functions dramatically increases what we can do with R.
Learning about R’s programming capabilities is an important step in gaining
facility with functions.

7.1 R Functions
Data on the yield (pounds per acre) of two types of corn seeds (regular and
kiln dried) were collected. Each of the 11 plots of land was split into two
subplots, and one of the subplots was planted in regular corn while the other
was planted in kiln dried corn. These data were analyzed in a famous paper
authored by William Gosset. Here are the data.

u.corn = "https://www.finley-lab.com/files/data/corn.csv"
corn = read.csv(u.corn, header=TRUE)
corn

regular kiln_dried
1 1903 2009
2 1935 1915
3 1910 2011
4 2496 2463
5 2108 2180
6 1961 1925
7 2060 2122
8 1444 1482
9 1612 1542
10 1316 1443
11 1511 1535

147
148 7 Functions and Programming

A paired t test, or a confidence interval for the mean difference, may be used
to assess the difference in yield between the two varieties. Of course R has
a function t.test that performs a paired t test and computes a confidence
interval, but we will perform the test without using that function. We will
focus for now on testing the hypotheses H0 : µd = 0 versus Ha : µd = 6 0 and
on a two-sided confidence interval for µd . Here µd represents the population
mean difference.
The paired t statistic is defined by
d
t= √ (7.1)
Sd / n

where d is the mean of the differences, Sd is the standard deviation of the

differences, and n is the sample size. The p-value is twice the area to the right
of |tobs |, where tobs is the observed t statistic, and a confidence interval is given
by √
d ± t∗ (Sd / n). (7.2)

Here t∗ is an appropriate quantile of a t distribution with n − 1 degrees of

freedom.

dbar <- mean(corn$kiln_dried - corn$regular)

n <- length(corn$regular)
S_d <- sd(corn$kiln_dried - corn$regular)
t_obs <- dbar/(S_d/sqrt(n))
t_obs

[1] 1.69

pval <- 2*(1 - pt(abs(t_obs), n-1))

pval

[1] 0.1218

margin <- qt(0.975, n-1)*(S_d/sqrt(n))

lcl <- dbar - margin
ucl <- dbar + margin
lcl

[1] -10.73

ucl

[1] 78.18
 7.1 R Functions 149

With a few lines of R code we have calculated the t statistic, the p-value, and
the confidence interval. Since paired t tests are pretty common, however, it
would be helpful to automate this procedure. One obvious reason is to save
time and effort, but another important reason is to avoid mistakes. It would
be easy to make a mistake (e.g., using n instead of n − 1 as the degrees of
freedom) when repeating the above computations.
Here is a first basic function which automates the computation.

paired_t <- function(x1, x2){

n <- length(x1)
dbar <- mean(x1 - x2)
s_d <- sd(x1 - x2)
tstat <- dbar/(s_d/sqrt(n))
pval <- 2*(1 - pt(abs(tstat), n-1))
margin <- qt(0.975, n-1)*s_d/sqrt(n)
lcl <- dbar - margin
ucl <- dbar + margin
return(list(tstat = tstat, pval = pval, lcl=lcl, ucl=ucl))
}

And here is the function in action

paired_t(x1 = corn$kiln_dried, x2 = corn$regular)

$tstat
[1] 1.69

$pval
[1] 0.1218

$lcl
[1] -10.73

$ucl
[1] 78.18
An explanation and comments on the function are in order.
• paired_t <- function(x1, x2) assigns a function of two variables, x1 and
x2, to an R object called paired_t.
• The compound expression, i.e., the code that makes up the body of the
function, is enclosed in curly braces {}.
• return(list(tstat = tstat, pval = pval, lcl=lcl, ucl=ucl)) indi-
cates the object(s) returned by the function. In this case the function returns
a list with four components.
 150 7 Functions and Programming

• The body of the function basically mimics the computations required to

compute the t statistic, the p-value, and the confidence interval.
• Several objects such as n and dbar are created in the body of the function.
These objects are NOT available outside the function. We will discuss this
further when we cover environments and scope in R.
Our function has automated the basic calculations. But it is still somewhat
limited in usefulness. For example, it only computes a 95% confidence interval,
while a user may want a different confidence level. And the function only
performs a two-sided test, while a user may want a one-sided procedure. We
modify the function slightly to allow the user to specify the confidence level
next.

paired_t <- function(x1, x2, cl = 0.95){

n <- length(x1)
dbar <- mean(x1 - x2)
s_d <- sd(x1 - x2)
tstat <- dbar/(s_d/sqrt(n))
pval <- 2*(1 - pt(abs(tstat), n-1))
pctile <- 1 - (1 - cl)/2
margin <- qt(pctile, n-1)*s_d/sqrt(n)
lcl <- dbar - margin
ucl <- dbar + margin
return(list(tstat = tstat, pval = pval, lcl = lcl, ucl=ucl))
}

paired_t(corn$kiln_dried, corn$regular)

$tstat
[1] 1.69

$pval
[1] 0.1218

$lcl
[1] -10.73

$ucl
[1] 78.18

paired_t(corn$kiln_dried, corn$regular, cl = 0.99)

$tstat
[1] 1.69
7.1 R Functions 151

$pval
[1] 0.1218

$lcl
[1] -29.5

$ucl
[1] 96.96
Two things to note. First, arguments do not have to be named. So
paired_t(corn$kiln_dried, corn$regular)
and
paired_t(x1 = corn$kiln_dried, x2 = corn$regular)
are equivalent. But we need to be careful if we do not name arguments because
then we have to know the ordering of the arguments in the function declaration.
Second, in the declaration of the function, the third argument cl was given a
default value of 0.95. If a user does not specify a value for cl it will silently
be set to 0.95. But of course a user can override this, as we did in
paired_t(corn$kiln_dried, corn$regular, cl = 0.99)

7.1.1 Practice Problem

Like all things in R, getting the hang of writing functions just requires practice.
Create a simple function called FtoK that is given a temperature in Farenheit
and converts it to Kelvin using the following formula

5
K = (F − 32) ∗ + 273.15
9
You should get the following if your function is correct

FtoK(80)

[1] 299.8

7.1.2 Creating Functions

Creating very short functions at the command prompt is a reasonable strategy.

For longer functions, one option is to write the function in a script and then
152 7 Functions and Programming

submit the whole function. Or a function can be written in any text editor,
saved as a plain text file (possibly with a .r extension), and then read into R
using the source() function.

7.2 Programming: Conditional Statements

The paired_t function is somewhat useful, but could be improved in several

ways. For example, consider the following:

paired_t(1:5, 1:4)

Warning in x1 - x2: longer object length is not a

multiple of shorter object length

Warning in x1 - x2: longer object length is not a

multiple of shorter object length
$tstat
[1] 1

$pval
[1] 0.3739

$lcl
[1] -1.421

$ucl
[1] 3.021
The user specified data had different numbers of observations in x1 and x2,
which of course can’t be data tested by a paired t test. Rather than stopping
and letting the user know that this is a problem, the function continued and
produced meaningless output.
Also, the function as written only allows testing against a two-sided alternative
hypothesis, and it would be good to allow one-sided alternatives.
First we will address some checks on arguments specified by the user. For this
we will use an if() function and a stop() function.

paired_t <- function(x1, x2, cl = 0.95){

if(length(x1) != length(x2)){
 7.2 Programming: Conditional Statements 153

stop("The input vectors must have the same length")

}
n <- length(x1)
dbar <- mean(x1 - x2)
s_d <- sd(x1 - x2)
tstat <- dbar/(s_d/sqrt(n))
pval <- 2*(1 - pt(abs(tstat), n-1))
pctile <- 1 - (1 - cl)/2
margin <- qt(pctile, n-1)*s_d/sqrt(n)
lcl <- dbar - margin
ucl <- dbar + margin
return(list(tstat = tstat, pval = pval, lcl = lcl, ucl=ucl))
}
paired_t(1:5, 1:4)

Error in paired_t(1:5, 1:4): The input vectors must have the same length
The argument to the if() function is evaluated. If the argument returns TRUE
the ensuing code is executed. Otherwise, the ensuing code is skipped and the
rest of the function is evaluated. If a stop() function is executed, the function
is exited and the argument of stop() is printed.
To better understand and use if() statements, we need to understand com-
parison operators and logical operators.

7.2.1 Comparison and Logical Operators

We have made use of some of the comparison operators in R. These include

• Equal: ==
• Not equal: !=
• Greater than: >
• Less than: <
• Greater than or equal to: >=
• Less than or equal to: <=
Special care needs to be taken with the == and != operators because of how
numbers are represented on computers, see Section 7.3.
There are also three logical operators, with two variants of the “and” operator
and the “or” operator.
• and: Either & or &&
• or: Either | or ||
• not: !
154 7 Functions and Programming

The “double” operators && and || just examine the first element of the two
vectors, whereas the “single” operators & and | compare element by element.

c(FALSE, TRUE, FALSE) || c(TRUE, FALSE, FALSE)

[1] TRUE

c(FALSE, TRUE, FALSE) | c(TRUE, FALSE, FALSE)

[1] TRUE TRUE FALSE

c(FALSE, TRUE, FALSE) && c(TRUE, TRUE, FALSE)

[1] FALSE

c(FALSE, TRUE, FALSE) & c(TRUE, TRUE, FALSE)

[1] FALSE TRUE FALSE

We can use the logical operators to check whether a user-specified confidence
level is between 0 and 1.

paired_t <- function(x1, x2, cl = 0.95){

if(length(x1) != length(x2)){
stop("The input vectors must have the same length")
}
if(cl <= 0 || cl >= 1){
stop("The confidence level must be between 0 and 1")
}
n <- length(x1)
dbar <- mean(x1 - x2)
s_d <- sd(x1 - x2)
tstat <- dbar/(s_d/sqrt(n))
pval <- 2*(1 - pt(abs(tstat), n-1))
pctile <- 1 - (1 - cl)/2
margin <- qt(pctile, n-1)*s_d/sqrt(n)
lcl <- dbar - margin
ucl <- dbar + margin
return(list(tstat = tstat, pval = pval, lcl = lcl, ucl=ucl))
}
paired_t(1:5, 2:6, cl=15)

Error in paired_t(1:5, 2:6, cl = 15): The confidence level must be between 0 and 1
7.2 Programming: Conditional Statements 155

7.2.2 If else statements

The if() statement we have used so far has the form

if (condition) {
expression
}
Often we want to evaluate one expression if the condition is true, and evaluate
a different expression if the condition is false. That is accomplished by the if
else statement. Here we determine whether a number is positive, negative, or
zero.

Sign <- function(x){

if(x < 0){
print("the number is negative")
}else if(x > 0){
print("the number is positive")
}else{
print("the number is zero")
}
}
Sign(3)

[1] "the number is positive"

Sign(-3)

[1] "the number is negative"

Sign(0)

[1] "the number is zero"

Notice the “different expression” for the first if statement was itself an if
statement.
Next we modify the paired_t function to allow two-sided and one-sided
alternatives.

paired_t <- function(x1, x2, cl = 0.95, alternative="not.equal"){

if(length(x1) != length(x2)){
stop("The input vectors must be of the same length")
}
156 7 Functions and Programming

if(cl <= 0 || cl >= 1){

stop("The confidence level must be between 0 and 1")
}
n <- length(x1)
dbar <- mean(x1 - x2)
s_d <- sd(x1 - x2)
tstat <- dbar/(s_d/sqrt(n))
if(alternative == "not.equal"){
pval <- 2*(1 - pt(abs(tstat), n-1))
pctile <- 1 - (1 - cl)/2
margin <- qt(pctile, n-1)*s_d/sqrt(n)
lcl <- dbar - margin
ucl <- dbar + margin
}else if(alternative == "greater"){
pval <- 1 - pt(tstat, n-1)
margin <- qt(cl, n-1)*s_d/sqrt(n)
lcl <- dbar - margin
ucl <- Inf
} else if(alternative == "less"){
pval <- pt(tstat, n-1)
margin <- qt(cl, n-1)*s_d/sqrt(n)
lcl <- -Inf
ucl <- dbar + margin
}

return(list(tstat = tstat, pval = pval, lcl=lcl, ucl=ucl))

}
paired_t(corn$kiln_dried, corn$regular)

$tstat
[1] 1.69

$pval
[1] 0.1218

$lcl
[1] -10.73

$ucl
[1] 78.18

paired_t(corn$kiln_dried, corn$regular, alternative = "less")

7.3 Computer Arithmetic 157

$tstat
[1] 1.69

$pval
[1] 0.9391

$lcl
[1] -Inf

$ucl
[1] 69.89

paired_t(corn$kiln_dried, corn$regular, alternative = "greater")

$tstat
[1] 1.69

$pval
[1] 0.06091

$lcl
[1] -2.434

$ucl
[1] Inf

7.3 Computer Arithmetic

Like most software, R does not perform exact arithmetic. Rather, R follows
the IEEE 754 floating point standards. This can have profound effects on
how computational algorithms are implemented, but is also important when
considering things like comparisons.
Note first that computer arithmetic does not follow some of the rules of ordinary
arithmetic. For example, it is not associative.

2ˆ-30

[1] 0.0000000009313
158 7 Functions and Programming

2ˆ-30 + (2ˆ30 - 2ˆ30)

[1] 0.0000000009313

(2ˆ-30 + 2ˆ30) - 2ˆ30

[1] 0
Computer arithmetic is not exact.

1.5 - 1.4

[1] 0.1

1.5 - 1.4 == 0.1

[1] FALSE

(1.5 - 1.4) - 0.1

[1] 0.00000000000000008327
So for example an if statement that uses an equality test may not give the
expected answer. One way to avoid this problem is to test “near equality” using
all.equal(). The function takes as arguments two objects to be compared,
and a tolerance. If the objects are within the tolerance of each other, the
function returns TRUE. The tolerance has a default value of about 1.5 × 10−8 ,
which works well in many cases.

all.equal((1.5 - 1.4), 0.1)

[1] TRUE

7.4 Loops

Loops are an important component of any programming language, including

R. Vectorized calculations and functions such as apply() make loops a bit less
central to R than to many other languages, but an understanding of the three
looping structures in R is still quite important.
7.4 Loops 159

We will investigate loops in the context of computing what is sometimes called

the “machine epsilon.” Because of the inexact representation of numbers in R
(and other languages) sometimes R cannot distinguish between the numbers 1
and |1 + x| for small values of x. The smallest value of x such that 1 and
|1+x| are not declared equal is the machine epsilon.

1 == 1+10ˆ-4

[1] FALSE

1 == 1 + 10ˆ-50

[1] TRUE
Clearly the machine epsilon is somewhere between 10−4 and 10−50 . How can we
find its value exactly? Since floating point numbers use a binary representation,
we know that the machine epsilon will be equal to 1/2k for some value of k.
So to find the machine epsilon, we can keep testing whether 1 and 1 + 1/2k
are equal, until we find a value k where the two are equal. Then the machine
epsilon will be 1/2k−1 , since it is the smallest value for which the two are NOT
equal.

1 == 1+1/2

[1] FALSE

1 == 1 + 1/2ˆ2

[1] FALSE

1 == 1 + 1/2ˆ3

[1] FALSE
Testing by hand like this gets tedious quickly. A loop can automate the process.
We will do this with two R loop types, repeat and while.

7.4.1 A Repeat Loop

A repeat loop just repeats a given expression over and over again until a
break statement is encountered.
160 7 Functions and Programming

k <- 1
repeat{
if(1 == 1+1/2ˆk){
break
}else{
k <- k+1
}
}
k

[1] 53

1/2ˆ(k-1)

[1] 0.000000000000000222
This code initializes k at 1. The body of the loop initially checks whether 1
and 1 + 1/2k are equal. If they are equal, the break statement is executed and
control is transferred outside the loop. If they are not equal, k is increased by
1, and we return to the beginning of the top of the body of the loop.

7.4.2 A While Loop

A while loop has the form

while (condition) {
expression
}
As long as the condition is TRUE the expression is evaluated. Once the
condition is FALSE control is transferred outside the loop.

k <- 1
while(1 != 1+1/2ˆk){
k <- k+1
}
k

[1] 53

1/2ˆ(k-1)

[1] 0.000000000000000222
7.4 Loops 161

7.4.3 A For Loop

A for loop has the form

for (variable in vector) {
expression
}
The for loop sets the variable equal to each element of the vector in
succession, and evaluates the expression each time. Here are two different
ways to use a for loop to calculate the sum of the elements in a vector.

x <- 1:10
S <- 0
for(i in 1:length(x)){
S <- S + x[i]
}
S

[1] 55

S = 0
for(value in x){
S = S + value
}
S

[1] 55
In the first case we loop over the positions of the vector elements, while in the
second case we loop over the elements themselves.

7.4.4 Practice Problem

Often when students initially learn about if() statements and for() loops,
they use them more often than is necessary, leading to over complications and
often slower, less sophisticated code. Many simple if() statements can be
accomplished using logical subsetting and vectorization. Using the ideas you
used above, and what you learned in Chapters 4 and 6, we can use logical
subsetting to replace the simplest of if() statements. Rewrite the following
if() statement and for() loop using logical subsetting.

a <- rnorm(1000)
for (i in 1:length(a)) {
162 7 Functions and Programming

if (a[i] > 1) {
a[i] <- 0
}
}

7.5 Efficiency Considerations

In many contexts R and modern computers are fast enough that the user does
not need to worry about writing efficient code. There are a few simple ways to
write efficient code that are easy enough, and provide enough speed-up, that
they are worth following as often as possible. The R function system.time()
reports how long a set of R code takes to execute, and we will use this function
to compare different ways to accomplish objectives in R.

7.5.1 Growing Objects

Consider two ways to create a sequence of integers from 1 to n, implemented

in functions f1 and f2.

1. Start with a zero-length vector, and let it grow:

f1 <- function(n){
x <- numeric(0)
for(i in 1:n){
x <- c(x,i)
}
x
}

2. Start with a vector of length n and fill in the values:

f2 <- function(n){
x <- numeric(n)
for(i in 1:n){
x[i] <- i
}
7.5 Efficiency Considerations 163

x
}

Here are the two functions in action, with n = 100000.

n <- 100000
system.time(f1(n))

user system elapsed

10.997 0.028 11.026

system.time(f2(n))

user system elapsed

0.007 0.000 0.007
It is much more efficient to start with a full-length vector and then fill in
values.1
Of course another way to create a vector of integers from 1 to n is to use 1:n.
Let’s see how fast this is.

system.time(1:n)

user system elapsed

0.001 0.000 0.005

n <- 1000000
system.time(1:n)

user system elapsed

0 0 0
For n = 100000 this is so fast the system time is very close to zero. Even
when n is 1000000 the time is very small. So another important lesson is to
use built-in R functions whenever possible, since they have had substantial
development focused on efficiency, correctness, etc.
1 Roughly speaking, the first option is slower because each time the vector is increased in

size, R must resize the vector and re-allocate memory.

164 7 Functions and Programming

7.5.2 Vectorization

Next consider calculating the sum of the squared entries in each column of a
matrix. For example with the matrix M ,

1 2 3
 
M= ,
4 5 6

the sums would be 12 + 42 = 17, 22 + 52 = 29, and 32 + 62 = 45. One possibility

is to have an outer loop that traverses the columns and an inner loop that
traverses the rows within a column, squaring the entries and adding them
together.

test_matrix <- matrix(1:6, byrow=TRUE, nrow=2)

test_matrix

[,1] [,2] [,3]

[1,] 1 2 3
[2,] 4 5 6

ss1 <- function(M){

n <- dim(M)[1]
m <- dim(M)[2]
out <- rep(0,m)
for(j in 1:m){
for(i in 1:n){
out[j] <- out[j] + M[i,j]ˆ2
}
}
return(out)
}
ss1(test_matrix)

[1] 17 29 45
Another possibility eliminates the inner loop, using the sum() function to
compute the sum of the squared entries in the column directly.

ss2 <- function(M){

m <- dim(M)[2]
out <- numeric(m)
for(j in 1:m){
out[j] <- sum(M[,j]ˆ2)
}
7.6 Efficiency Considerations 165

return(out)
}
ss2(test_matrix)

[1] 17 29 45
A third possibility uses the colSums() function.

ss3 <- function(M){

out <- colSums(Mˆ2)
return(out)
}
ss3(test_matrix)

[1] 17 29 45
Here is a speed comparison, using a 1000 × 10000 matrix.

mm <- matrix(1:10000000, byrow=TRUE, nrow=1000)

system.time(ss1(mm))

user system elapsed

0.879 0.000 0.880

system.time(ss2(mm))

user system elapsed

0.075 0.028 0.103

system.time(ss3(mm))

user system elapsed

0.207 0.016 0.223

rm(mm)
166 7 Functions and Programming

7.6 More on Functions

Understanding functions deeply requires a careful study of R’s scoping rules,
as well as a good understanding of environments in R. That’s beyond the scope
of this book, but we will briefly discuss some issues that are most salient. For
a more in-depth treatment, see “Advanced R” by Hadley Wickham, especially
the chapters on functions and environments.

7.7 Calling Functions

When using a function, the functions arguments can be specified in three ways:

1. By the full name of the argument.

2. By the position of the argument.
3. By a partial name of the argument.

tmp_function <- function(first.arg, second.arg, third.arg, fourth.arg){

return(c(first.arg, second.arg, third.arg, fourth.arg))
}
tmp_function(34, 15, third.arg = 11, fou = 99)

[1] 34 15 11 99
Positional matching of arguments is convenient, but should be used carefully,
and probably limited to the first few, and most commonly used, arguments
in a function. Partial does have pitfalls. A partially specified argument must
unambiguously match exactly one argument—a requirement that’s not met
below.

tmp_function <- function(first.arg, fourth.arg){

return(c(first.arg, fourth.arg))
}
tmp_function(1, f=2)

Error in tmp_function(1, f = 2): argument 2 matches multiple formal arguments

7.7 Calling Functions 167

7.7.1 The ... argument

In defining a function, a special argument denoted by ... can be used. Some-

times this is called the “ellipsis” argument, sometimes the “three dot” argu-
ment, sometimes the “dot dot dot” argument, etc. The R language defini-
tion https://cran.r-project.org/doc/manuals/r-release/R-lang.html
describes the argument in this way:

The special type of argument ‘. . . ’ can contain any number

of supplied arguments. It is used for a variety of purposes. It
allows you to write a function that takes an arbitrary number
of arguments. It can be used to absorb some arguments into an
intermediate function which can then be extracted by functions
called subsequently.

Consider for example the sum() function.

sum(1:5)

[1] 15

sum(1:5, c(3,4,90))

[1] 112

sum(1,2,3,c(3,4,90), 1:5)

[1] 118
Think about writing such a function. There is no way to predict in advance
the number of arguments a user might specify. So the function is defined with
... as the first argument:

sum

function (..., na.rm = FALSE) .Primitive("sum")

This is true of many commonly-used functions in R such as c() among others.
Next, consider a function that calls another function in its body. For example,
168 7 Functions and Programming

suppose that a collaborator always supplies comma delimited files that have
five lines of description, followed by a line containing variable names, followed
by the data. You are tired of having to specify skip = 5, header = TRUE, and
sep = "," to read.table() and want to create a function my.read() which
uses these as defaults.

my.read <- function(file, header=TRUE, sep = ",", skip = 5, ...){

read.table(file = file, header = header, sep = sep, skip = skip, ...)
}

The ... in the definition of my.read() allows the user to specify other argu-
ments, for example, stringsAsFactors = FALSE. These will be passed on to
the read.table() function. In fact, that is how read.csv() is defined.

read.csv

function (file, header = TRUE, sep = ",", quote = "\"", dec = ".",
fill = TRUE, comment.char = "", ...)
read.table(file = file, header = header, sep = sep, quote = quote,
dec = dec, fill = fill, comment.char = comment.char, ...)
<bytecode: 0x564843e4e570>
<environment: namespace:utils>

7.7.2 Lazy Evaluation

Arguments to R functions are not evaluated until they are needed, sometimes
called lazy evaluation.

f <- function(a,b){
print(aˆ2)
print(aˆ3)
print(a*b)
}
f(a=3, b=2)

[1] 9
[1] 27
[1] 6

f(a=3)

[1] 9
[1] 27
7.8 Exercises 169

Error in print(a * b): argument "b" is missing, with no default

The first call specified both of the arguments a and b, and produced the
expected output. In the second call the argument b was not specified. Since
it was not needed until the third print statement, R happily executed the
first two print statements, and only reported an error in the third statement,
when b was needed to compute a*b.

f <- function(a,b = aˆ3){

return(a*b)
}
f(2)

[1] 16

f(2,10)

[1] 20
In the first call, since b was not specified, it was computed as aˆ3. In the
second call, b was specified, and the specified value was used.

7.8 Exercises
Exercise 10 Learning objectives: translate statistical notation into coded
functions; learn about tools for checking validity of function arguments; practice
writing functions that return multiple objects.
 8
Spatial Data Visualization and Analysis

8.1 Overview

Recall, a data structure is a format for organizing and storing data. The struc-
ture is designed so that data can be accessed and worked with in specific ways.
Statistical software and programming languages have methods (or functions)
designed to operate on different kinds of data structures.
This chapter focuses on spatial data structures and some of the R functions that
work with these data. Spatial data comprise values associated with locations,
such as temperature data at a given latitude, longitude, and perhaps elevation.
Spatial data are typically organized into vector or raster data types. (See
Figure 8.1).
• Vector data represent features such as discrete points, lines, and polygons.
• Raster data represent surfaces as a rectangular matrix of square cells or
pixels.
Whether or not you use vector or raster data depends on the type of problem,
the type of maps you need, and the data source. Each data structure has
strengths and weaknesses in terms of functionality and representation. As you
gain more experience working with spatial data, you will be able to determine
which structure to use for a particular application.
There is a large set of R packages available for working with spatial (and space-
time) data. These packages are described in the Cran Task View: Analysis of
Spatial Data1 . The CRAN task view attempts to organize the various packages
into categories, such as Handling spatial data, Reading and writing spatial
data, Visualization, and Disease mapping and areal data analysis, so users can
quickly identify package options given their project needs.
Exploring the extent of the excellent spatial data tools available in R is beyond
the scope of this book. Rather, we would point you to subject texts like Applied
Spatial Data Analysis with R by Bivand et al. (2013) (available for free via the
MSU library system), and numerous online tutorials on pretty much any aspect
1 https://CRAN.R-project.org/view=Spatial

171
 172 8 Spatial Data Visualization and Analysis

FIGURE 8.1: Raster/Vector Comparison [@imageRaster]

of spatial data analysis with R. These tools make R a full-blown Geographic

Information System2 (GIS) capable of spatial data manipulation and analysis
on par with commercial GIS systems such as ESRI’s ArcGIS3 .

8.1.1 Some Spatial Data Packages

This chapter will focus on a few R packages for manipulating and visualizing
spatial data. Specifically we will touch on the following packages
• sp: spatial data structures and methods
• rgdal: interface to the C/C++ spatial data Geospatial Data Abstraction
Library
• ggmap: extends ggplot2 language to handle spatial data
• leaflet: generates dynamic online maps
2 https://en.wikipedia.org/wiki/Geographic_information_system
3 http://www.esri.com/arcgis/about-arcgis
8.3 Motivating Data 173

8.2 Motivating Data

We motivate the topics introduced in this chapter using some forestry data
from the Penobscot Experimental Forest4 (PEF) located in Maine. The PEF
is a long-term experimental forest that is used to understand the effects of
silviculture (i.e., science of tending trees) treatments on forest growth and
composition. The PEF is divided into non-overlapping management units that
receive different harvesting treatments. Within each management unit is a
series of observation point locations (called forest inventory plots) where forest
variables have been measured. Ultimately, we want to summarize the inventory
plots measurements by management unit and map the results.

8.3 Reading Spatial Data into R

Spatial data come in a variety of file formats. Examples of popular vector
file formats for points, lines, and polygons, include ESRI’s shapefile5 and
open standard GeoJSON6 . Common raster file formats include GeoTIFF7 and
netCDF89 .
The rgdal function readOGR will read a large variety of vector data file formats
(there is also a writeOGR() for writing vector data files). Raster data file formats
can be read using the rgdal function readGDAL (yup, also a writeGDAL()) or
read functions in the raster package. All of these functions automatically cast
the data into an appropriate R spatial data object (i.e., data structure), which
are defined in the sp or raster packages. Table 8.1 provides an abbreviated
list of these R spatial objects10 . The Without attributes column gives the sp
package’s spatial data object classes for points, lines, polygons, and raster
pixels that do not have data associated with the spatial objects (i.e., without
4 https://www.nrs.fs.fed.us/ef/locations/me/penobscot%7D%7BPenobscot%20Exper

imental%20Forest
5 https://en.wikipedia.org/wiki/Shapefile
6 https://en.wikipedia.org/wiki/GeoJSON
7 https://en.wikipedia.org/wiki/GeoTIFF
8 https://en.wikipedia.org/wiki/NetCDF
9 A longer list of spatial data file formats is available at https://en.wikipedia.org/wik

i/GIS_file_formats.
10 A more complete list of the sp package’s spatial data classes and methods is detailed

in the package’s vignette https://cran.r-project.org/web/packages/sp/vignettes/intr

o_sp.pdf.
174 8 Spatial Data Visualization and Analysis

TABLE 8.1: An abbreviated list of ‘sp‘ and ‘raster‘ data objects and associ-
ated classes for the fundamental spatial data types

Without Attributes With Attributes

Polygons SpatialPolygons SpatialPolygonsDataFrame
Points SpatialPoints SpatialPointsDataFrame
Lines SpatialLines SpatialLinesDataFrame
Raster SpatialGrid SpatialGridDataFrame
Raster SpatialPixels SpatialPixelsDataFrame
Raster RasterLayer
Raster RasterBrick
Raster RasterStack

attributes in GIS speak). DataFrame is appended to the object class name

once data, in the form of variables, are added to the spatial object.
You can create your own spatial data objects in R. Below, for example, we
create a SpatialPoints object consisting of four points. Then add some data
to the points to make it a SpatialPointsDataFrame.

library(sp)
library(dplyr)

x <- c(3,2,5,6)
y <- c(2,5,6,7)

coords <- cbind(x, y)

sp.pts <- SpatialPoints(coords)

class(sp.pts)

[1] "SpatialPoints"
attr(,"package")
[1] "sp"

some.data <- data.frame(var.1 = c("a", "b", "c", "d"), var.2 = 1:4)

sp.pts.df <- SpatialPointsDataFrame(sp.pts, some.data)

class(sp.pts.df)

[1] "SpatialPointsDataFrame"
8.3 Reading Spatial Data into R 175

attr(,"package")
[1] "sp"
If, for example, you already have a data frame that includes the spatial
coordinate columns and other variables, then you can promote it to a
SpatialPointsDataFrame by indicating which columns contain point coordi-
nates. You can extract or access the data frame associated with the spatial
object using @data. You can also access individual variables directly from the
spatial object using $ or by name or column number to the right of the comma
in [,] (analogues to accessing variables in a data frame).

df <- data.frame(x = c(3,2,5,6), y=c(2,5,6,7), var.1 = c("a", "b", "c", "d"), var.2 = 1:4)
class(df)

[1] "data.frame"

#promote to a SpatialPointsDataFrame
coordinates(df) <- ~x+y

class(df)

[1] "SpatialPointsDataFrame"
attr(,"package")
[1] "sp"

#access entire data frame

df@data

var.1 var.2
1 a 1
2 b 2
3 c 3
4 d 4

class(df@data)

[1] "data.frame"

#access columns directly

df$var.1

[1] "a" "b" "c" "d"

176 8 Spatial Data Visualization and Analysis

df[,c("var.1","var.2")]

coordinates var.1 var.2

1 (3, 2) a 1
2 (2, 5) b 2
3 (5, 6) c 3
4 (6, 7) d 4

df[,2]

coordinates var.2
1 (3, 2) 1
2 (2, 5) 2
3 (5, 6) 3
4 (6, 7) 4

#get the bounding box

bbox(df)

min max
x 2 6
y 2 7
Here, the data frame df is promoted to a SpatialPointsDataFrame by in-
dicating the column names that hold the longitude and latitude (i.e., x and
y respectively) using the coordinates function. Here too, the @data is used
to retrieve the data frame associated with the points. We also illustrate how
variables can be accessed directly from the spatial object. The bbox function
is used to return the bounding box that is defined by the spatial extent of the
point coordinates. The other spatial objects noted in Table 8.1 can be created,
and their data accessed, in a similar way11 .
More than often we find ourselves reading existing spatial data files into R.
The code below uses the downloader package to download all of the PEF
data we’ll use in this chapter. The data are compressed in a single zip file,
which is then extracted into the working directory using the unzip function.
A look into the PEF directory using list.files shows nine files12 . Those
named MU-bounds.* comprise the shapefile that holds the PEF’s management
unit boundaries in the form of polygons. Like other spatial data file formats,
11 This cheatsheet http://www.maths.lancs.ac.uk/~rowlings/Teaching/UseR2012/chea

tsheet.html written by Barry Rowlingson is a useful reference http://www.maths.lancs.

ac.uk/~rowlings/Teaching/UseR2012/cheatsheet.html
12 The list.files function does not read data into R; it simply prints the contents of a

directory.
8.3 Reading Spatial Data into R 177

shapefiles are made up of several different files (with different file extensions)
that are linked together to form a spatial data object. The plots.csv file
holds the spatial coordinates and other information about the PEF’s forest
inventory plots. The roads.* shapefile holds roads and trails in and around
the PEF.

library(downloader)

download("https://www.finley-lab.com/files/data/PEF.zip",
destfile="./PEF.zip", mode="wb")

unzip("PEF.zip", exdir = ".")

list.files("PEF")

[1] "MU-bounds.dbf" "MU-bounds.prj" "MU-bounds.qpj"

[4] "MU-bounds.shp" "MU-bounds.shx" "plots.csv"
[7] "plots.dbf" "plots.prj" "plots.qpj"
[10] "plots.shp" "plots.shx" "roads.dbf"
[13] "roads.prj" "roads.shp" "roads.shx"
Next we read the MU-bounds shapefile into R using readOGR()13 and explore
the resulting mu object. Notice that when we read a shapefile into R, we do not
include a file extension with the shapefile name because a shapefile is always
composed of multiple files.

library(rgdal)

rgdal: version: 1.5-12, (SVN revision 1018)

Geospatial Data Abstraction Library extensions to R successfully loaded
Loaded GDAL runtime: GDAL 2.2.3, released 2017/11/20
Path to GDAL shared files: /usr/share/gdal/2.2
GDAL binary built with GEOS: TRUE
Loaded PROJ runtime: Rel. 4.9.3, 15 August 2016, [PJ_VERSION: 493]
Path to PROJ shared files: (autodetected)
Linking to sp version:1.4-2

mu <- readOGR("PEF", "MU-bounds")

OGR data source with driver: ESRI Shapefile

Source: "/home/jeffdoser/Dropbox/teaching/for875-20/textbook/PEF", layer: "MU-bounds"
13 The authors of the rgdal library decided to have some information about the version

of GDAL and other software specifics be printed when the library is loaded. Don’t let it
distract you.
178 8 Spatial Data Visualization and Analysis

with 40 features
It has 1 fields
When called, the readOGR function provides a bit of information about the
object being read in. Here, we see that it read the MU-bounds shapefile from
PEF directory and the shapefile had 40 features (i.e., polygons) and 1 field
(i.e., field is synonymous with column or variable in the data frame).
You can think of the resulting mu object as a data frame where each row corre-
sponds to a polygon and each column holds information about the polygon14 .
More specifically, the mu object is a SpatialPolygonsDataFrame.

class(mu)

[1] "SpatialPolygonsDataFrame"
attr(,"package")
[1] "sp"
As illustrated using the made-up point data in the example above, you can
access the data frame associated with the polygons using @data.

class(mu@data)

[1] "data.frame"

dim(mu@data)

[1] 40 1

head(mu@data)

mu_id
0 C15
1 C17
2 C16
3 C27
4 U18
5 U31
Above, a call to class() confirms we have accessed the data frame, dim()
shows there are 40 rows (one row for each polygon) and one column, and
head() shows the first six values of the column named mu_id. The mu_id
14 Much of the actual spatial information is hidden from you in other parts of the data

structure, but is available if you ask nicely for it (see subsequent sections).
8.4 Coordinate Reference Systems 179

values are unique identifiers for each management unit polygon across the
PEF.

8.4 Coordinate Reference Systems

One of the more challenging aspects of working with spatial data is getting
used to the idea of a coordinate reference system. A coordinate reference system
(CRS) is a system that uses one or more numbers, or coordinates, to uniquely
determine the position of a point or other geometric element (e.g., line, polygon,
raster). For spatial data, there are two common coordinate systems:

1. Spherical coordinate system, such as latitude-longitude, often re-

ferred to as a geographic coordinate system.
2. Projected coordinate system based on a map projection, which is a
systematic transformation of the latitudes and longitudes that aims
to minimize distortion occurring from projecting maps of the earth’s
spherical surface onto a two-dimensional Cartesian coordinate plane.
Projected coordinate systems are sometimes referred to as map
projections.

There are numerous map projections15 . One of the more frustrating parts of
working with spatial data is that it seems like each data source you find offers
its data in a different map projection and hence you spend a great deal of time
reprojecting (i.e., transforming from one CRS to another) data into a common
CRS such that they overlay correctly. Reprojecting is accomplished using the
sp package’s spTransform function as demonstrated in Section 8.5.
In R, a spatial object’s CRS is accessed via the sp package proj4string
function. The code below shows the current projection of mu.

proj4string(mu)

[1] "+proj=utm +zone=19 +datum=NAD83 +units=m +no_defs "

The cryptic looking string returned by proj4string() is a set of directives
understood by the proj.416 C library, which is part of sp, and used to map
geographic longitude and latitude coordinates into the projected Cartesian
15 See partial list of map projections at https://en.wikipedia.org/wiki/List/_of/_map

/_projections. See a humorous discussion of map projections at http://brilliantmaps.

com/xkcd/.
16 http://proj4.org/
180 8 Spatial Data Visualization and Analysis

coordinates. This CRS tells us the mu object is in Universal Transverse Mercator

(UTM)17 zone 19 coordinate system.18

8.5 Illustration using ggmap

Let’s begin by making a map of PEF management unit boundaries over top
of a satellite image using the ggmap package. Given an address, location, or
bounding box, the ggmap package’s get_map function will query Google Maps,
OpenStreetMap, Stamen Maps, or Naver Map servers for a user-specified map
type. The get_map function requires the location or bounding box coordinates
be in a geographic coordinate system (i.e., latitude-longitude). This means
we need to reproject mu from UTM zone 19 to latitude-longitude geographic
coordinates, which is defined by the '"proj=longlat +datum=WGS84"' proj.4
string. As seen below, the first argument in spTransform function is the spatial
object to reproject and the second argument is a CRS object created by passing
a proj.4 string into the CRS function.

mu <- spTransform(mu, CRS("+proj=longlat +datum=WGS84"))

proj4string(mu)

[1] "+proj=longlat +datum=WGS84 +ellps=WGS84 +towgs84=0,0,0"

Unfortunately, we cannot just feed the SpatialPolygonsDataFrame mu into
ggplot (perhaps some day soon this will possible). Rather, we need to first
convert the SpatialPolygonsDataFrame into a specially formatted data frame
using the fortify function that is part of the ggplot2 package19 . The fortify
function will also need a unique identifier for each polygon specified using the
region argument, which for mu is the mu_id.

library(ggmap)
mu.f <- fortify(mu, region="mu_id")
head(mu.f)

long lat order hole piece id group

1 -68.62 44.86 1 FALSE 1 C12 C12.1
17 https://en.wikipedia.org/wiki/Universal_Transverse_Mercator_coordinate_sys

tem
18 If you start dealing with a lot of spatial data and reprojecting, http://spatialreferen

ce.org is an excellent resources for finding and specifying coordinate reference systems.
19 ggmap depends on ggplot2 so ggplot2 will be automatically loaded when you call

library(ggmap).
8.5 Illustration using ggmap 181

2 -68.62 44.86 2 FALSE 1 C12 C12.1

3 -68.62 44.86 3 FALSE 1 C12 C12.1
4 -68.62 44.86 4 FALSE 1 C12 C12.1
5 -68.62 44.86 5 FALSE 1 C12 C12.1
6 -68.62 44.86 6 FALSE 1 C12 C12.1
Notice the id column in the fortified version of mu holds each polygon’s mu_id
value (this will be important later when we link data to the polygons).
Next, we query the satellite imagery used to underlay the management units
(we’ll generally refer to this underlying map as the basemap). As of October
2018, Google requires you to set up a Google API account to download imagery.
This is free, but it does require a credit card to obtain the API Key that is
required to make the ggmap package work. Here I provide you with an API key
for a project I created for this class that should allow you to run the following
function if you desire. If you are interested in obtaining your own API key, see
the page here20 for learning about how to use Google maps web services.

register_google(key = "AIzaSyBPAwSY5x8vQqlnG-QwiCAWQW12U3CTLZY")
mu.bbox <- bbox(mu)
mu.centroid <- apply(mu.bbox, 1, mean)

basemap <- get_map(location=mu.centroid, zoom=14, maptype = "satellite", source = "google"

ggmap(basemap) +
geom_polygon(data=mu.f, aes(x = long, y = lat, group=group),
fill=NA, size=0.2, color="orange")

20 https://developers.google.com/maps/documentation/geocoding/get-api-key
182 8 Spatial Data Visualization and Analysis

44.87

lat 44.86

44.85

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

While the call to get_map can use a bounding box (mu.bbox in code above)
to define the basemap extent, I’ve found that it’s easier to pass it a center
point location (mu.centroid note the use of the apply function to find the
mean in the easting and northing extents) and zoom value to extract the
desired extent. The resulting map looks pretty good! Take a look at the
get_map function manual page and try different options for maptype (e.g.,
maptype="terrain").
Next we’ll add the forest inventory plots to the map. Begin by reading in
the PEF forest inventory plot data held in “plots.csv”. Recall, foresters have
measured forest variables at a set of locations (i.e., inventory plots) within
each management unit. The following statement reads these data and displays
the resulting data frame structure.

plots <- read.csv("PEF/plots.csv", stringsAsFactors=FALSE)

str(plots)

'data.frame': 451 obs. of 8 variables:

$ mu_id : chr "U10" "U10" "U10" "U10" ...
$ plot : int 11 13 21 22 23 24 31 32 33 34 ...
$ easting : num 529699 529777 529774 529814 529850 ...
$ northing : num 4966333 4966471 4966265 4966336 4966402 ...
$ biomass_mg_ha : num 96.3 115.7 121.6 72 122.3 ...
$ stems_ha : int 5453 2629 3385 7742 7980 10047 5039 5831 2505 7325 ...
$ diameter_cm : num 4.8 6.9 6.1 3.1 4.7 1.6 4.1 5.2 5.7 3.3 ...
8.5 Illustration using ggmap 183

$ basal_area_m2_ha: num 22 23.2 23 16.1 29.2 19.1 14.1 27.4 21.6 15 ...
In plots each row is a forest inventory plot and columns are:
• mu_id identifies the management unit within which the plot is located
• plot unique plot number within the management unit
• easting longitude coordinate in UTM zone 19
• northing latitude coordinate in UTM zone 19
• biomass_mg_ha tree biomass in metric ton (per hectare basis)
• stocking_stems_ha number of tree (per hectare basis)
• diameter_cm average tree diameter measured 130 cm up the tree trunk
• basal_area_m2_ha total cross-sectional area at 130 cm up the tree trunk
(per hectare basis)
There is nothing inherently spatial about this data structure—it is simply a
data frame. We make plots into a spatial object by identifying which columns
hold the coordinates. This is done below using the coordinates function,
which promotes the plots data frame to a SpatialPointsDataFrame.

coordinates(plots) <- ~easting+northing

class(plots)

[1] "SpatialPointsDataFrame"
attr(,"package")
[1] "sp"
Although plots is now a SpatialPointsDataFrame, it does not know to which
CRS the coordinates belong; hence, the NA when proj4string(plots) is called
below. As noted in the plots file description above, easting and northing
are in UTM zone 19. This CRS is set using the second call to proj4string
below.

proj4string(plots)

[1] NA

proj4string(plots) <- CRS("+proj=utm +zone=19 +datum=NAD83 +units=m

+no_defs +ellps=GRS80 +towgs84=0,0,0")

Now let’s reproject plots to share a common CRS with mu

plots <- spTransform(plots, CRS("+proj=longlat +datum=WGS84"))

Note, because mu is already in the projection we want for plots, we could

184 8 Spatial Data Visualization and Analysis

have replaced the second argument in the spTransform call above with
proj4string(mu) and saved some typing.
We’re now ready to add the forest inventory plots to the existing basemap
with management units. Specifically, let’s map the biomass_mg_ha variable to
show changes in biomass across the forest. No need to fortify plots, ggplot
is happy to take geom_point’s data argument as a data frame (although we
do need to convert plots from a SpatialPointsDataFrame to a data frame
using the as.data.frame function). Check out the scale_color_gradient
function in your favorite ggplot2 reference to understand how the color scale
is set.

ggmap(basemap) +
geom_polygon(data=mu.f, aes(x = long, y = lat, group=group),
fill=NA, size=0.2, color="orange") +
geom_point(data=as.data.frame(plots),
aes(x = easting, y = northing, color=biomass_mg_ha)) +
scale_color_gradient(low="white", high="darkblue") +
labs(color = "Biomass (mg/ha)")

44.87

44.86
Biomass (mg/ha)
200

150
lat

44.85 100

50

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

There is something subtle and easy to miss in the code above. Notice the aes
function arguments in geom_points take geographic longitude and latitude,
x and y respectively, from the points data frame (but recall easting and
northing were in UTM zone 19). This works because we applied spTransform
to reproject the points SpatialPointsDataFrame to geographic coordinates.
8.5 Illustration using ggmap 185

sp then replaces the values in easting and northing columns with the re-
projected coordinate values when converting a SpatialPointsDataFrame to
a data frame via as.data.frame().
Foresters use the inventory plot measurements to estimate forest variables
within management units, e.g., the average or total management unit biomass.
Next we’ll make a plot with management unit polygons colored by average
biomass_mg_ha.

mu.bio <- plots@data %>% group_by(mu_id) %>%

summarize(biomass_mu = mean(biomass_mg_ha))

`summarise()` ungrouping output (override with `.groups` argument)

print(mu.bio)

# A tibble: 33 x 2
mu_id biomass_mu
<chr> <dbl>
1 C12 124.
2 C15 49.9
3 C16 128.
4 C17 112.
5 C20 121.
6 C21 134.
7 C22 65.2
8 C23A 108.
9 C23B 153.
10 C24 126.
# ... with 23 more rows
Recall from Section 6.7.5 this one-liner can be read as “get the data frame
from plots’s SpatialPointsDataFrame then group by management unit then
make a new variable called biomass_mu that is the average of biomass_mg_ha
and assign it to the mu.bio tibble.”
The management unit specific biomass_mu can now be joined to the mu polygons
using the common mu_id value. Remember when we created the fortified version
of mu called mu.f? The fortify function region argument was mu_id which is
the id variable in the resulting mu.f. This id variable in mu.f can be linked to
the mu_id variable in mu.bio using dplyr’s left_join function as illustrated
below.

head(mu.f, n=2)
186 8 Spatial Data Visualization and Analysis

long lat order hole piece id group

1 -68.62 44.86 1 FALSE 1 C12 C12.1
2 -68.62 44.86 2 FALSE 1 C12 C12.1

mu.f <- left_join(mu.f, mu.bio, by = c('id' = 'mu_id'))

head(mu.f, n=2)

long lat order hole piece id group biomass_mu

1 -68.62 44.86 1 FALSE 1 C12 C12.1 123.7
2 -68.62 44.86 2 FALSE 1 C12 C12.1 123.7
The calls to head() show the first few rows of mu.f pre- and post-join. After the
join, mu.f includes biomass_mu, which is used used below for geom_polygon’s
fill argument to color the polygons accordingly.

ggmap(basemap) +
geom_polygon(data=mu.f, aes(x = long, y = lat,
group=group, fill=biomass_mu),
size=0.2, color="orange") +
scale_fill_gradient(low="white", high="darkblue",
na.value="transparent") +
labs(fill="Biomass (mg/ha)")

44.87

44.86
Biomass (mg/ha)

150
lat

100
44.85

50

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

Let’s add the roads and some more descriptive labels as a finishing touch. The
8.5 Illustration using ggmap 187

roads data include a variable called type that identifies the road type. To
color roads by type in the map, we need to join the roads data frame with
the fortified roads roads.f using the common variable id as a road segment
specific identifier. Then geom_path’s color argument gets this type variable
as a factor to create road-specific color. The default coloring of the roads blends
in too much with the polygon colors, so we manually set the road colors using
the scale_color_brewer function. The palette argument in this function
accepts a set of key words, e.g., "Dark2", that specify sets of diverging colors
chosen to make map object difference optimally distinct (see, the manual page
for scale_color_brewer, http://colorbrewer2.org, and blog here21 .)22

roads <- readOGR("PEF", "roads")

OGR data source with driver: ESRI Shapefile

Source: "/home/jeffdoser/Dropbox/teaching/for875-20/textbook/PEF", layer: "roads"
with 33 features
It has 2 fields

roads <- spTransform(roads, proj4string(mu))

roads.f <- fortify(roads, region="id")

roads.f <- left_join(roads.f, roads@data, by = c('id' = 'id'))

ggmap(basemap) +
geom_polygon(data=mu.f, aes(x = long, y = lat, group=group,
fill=biomass_mu),
size=0.2, color="orange") +
geom_path(data=roads.f, aes(x = long, y = lat,
group=group, color=factor(type))) +
scale_fill_gradient(low="white", high="darkblue",
na.value="transparent") +
scale_color_brewer(palette="Dark2") +
labs(fill="Biomass (mg/ha)", color="Road type", xlab="Longitude",
ylab="Latitude", title="PEF forest biomass")

Warning: Removed 616 row(s) containing missing values

(geom_path).
21 https://www.r-bloggers.com/r-using-rcolorbrewer-to-colour-your-figures-in-

r
22 Install the RColorBrewer package and run library(RColorBrewer);
display.brewer.all() to get a graphical list of available palettes.
188 8 Spatial Data Visualization and Analysis

PEF forest biomass

44.87
Road type
Gravel
Paved

44.86 Paved/Gravel
Trail
Winter
lat

44.85 Biomass (mg/ha)

150

100
44.84
50

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

The second, and more cryptic, of the two warnings from this code occurs
because some of the roads extend beyond the range of the map axes and are
removed (nothing to worry about).

8.6 Illustration using leaflet

Leaflet is one of the most popular open-source JavaScript libraries for interactive
maps. As noted on the official R leaflet website23 , it’s used by websites ranging
from The New York Times and The Washington Post to GitHub and Flickr,
as well as by GIS specialists like OpenStreetMap, Mapbox, and CartoDB.
The R leaflet website24 is an excellent resource to learn leaflet basics, and
should serve as a reference to gain a better understanding of the topics we
briefly explore below.
You create a leaflet map using these basic steps:

1. Create a map by calling leaflet()

2. Add data layers to the map using layer functions
23 https://rstudio.github.io/leaflet
24 https://rstudio.github.io/leaflet
8.6 Illustration using leaflet 189

such as, addTiles(), addMarkers(), addPolygons(),

addCircleMarkers(), addPolylines(), addRasterImage()
and other add... functions
3. Repeat step 2 to add more layers to the map
4. Print the map to display it

Here’s a brief example.

library(leaflet)

m <- leaflet() %>%

addTiles() %>% # Add default OpenStreetMap map tiles
addMarkers(lng=-84.482004, lat=42.727516,
popup="<b>Here I am!</b>") # Add a clickable marker
m # Print the map

There are a couple things to note in the code. First, we use the pipe operator
%>% just like in dplyr functions. Second, the popup argument in addMarkers()
takes standard HTML and clicking on the marker makes the text popup.
Third, the html version of this text provides the full interactive, dynamic
map, so we encourage you to read and interact with the html version of this
textbook for this section. The PDF document will simply display a static
version of this map and will not do justice to how awesome leaflet truly is!
As seen in the leaflet() call above, the various add... functions can take
longitude (i.e., lng) and latitude (i.e., lat). Alternatively, these functions can
extract the necessary spatial information from sp objects, e.g., Table 8.1, when
190 8 Spatial Data Visualization and Analysis

passed to the data argument (which greatly simplifies life compared with map
making using ggmap).

8.7 Subsetting Spatial Data

You can imagine that we might want to subset spatial objects to map specific
points, lines, or polygons that meet some criteria, or perhaps extract values
from polygons or raster surfaces at a set of points or geographic extent. These,
and similar types, of operations are easy in R (as long as all spatial objects
are in a common CRS). Recall from Chapter 4 how handy it is to subset data
structures, e.g., vectors and data frames, using the [] operator and logical
vectors? Well it’s just as easy to subset spatial objects, thanks to the authors
of sp, raster, and other spatial data packages.

8.7.1 Fetching and Cropping Data using raster

In order to motivate our exploration of spatial data subsetting and to illus-

trate some useful functionality of the raster package, let’s download some
elevation data for the PEF. The raster package has a rich set of functions
for manipulating raster data as well as functions for downloading data from
open source repositories. We’ll focus on the package’s getData function, which,
given a location in geographic longitude and latitude or location name, will
download data from GADM25 , Shuttle Radar Topography Mission26 , Global
Climate Data27 , and other sources commonly used in spatial data applications.
Let’s download SRTM surface elevation data for the PEF, check the resulting
object’s class and CRS, and display it using the raster package’s image
function along with the PEF forest inventory plots.

library(raster)

pef.centroid <- as.data.frame(plots) %>%

summarize(mu.x = mean(easting), mu.y = mean(northing))

srtm <- getData("SRTM", lon = pef.centroid[, 1], lat = pef.centroid[, 2])

25 http://www.gadm.org/%7D%7BGlobal%20Administrative%20Boundaries
26 https://www2.jpl.nasa.gov/srtm/%7D%7BShuttle%20Radar%20Topography%20Mission
27 http://www.worldclim.org/
8.7 Subsetting Spatial Data 191

class(srtm)

[1] "RasterLayer"
attr(,"package")
[1] "raster"

proj4string(srtm)

[1] "+proj=longlat +datum=WGS84 +ellps=WGS84 +towgs84=0,0,0"

image(srtm)
plot(plots, add = TRUE)
45
44
43
y

42
41
40

-70 -69 -68 -67 -66 -65

A few things to notice in the code above. First the getData function needs the
longitude lon and latitude lat to identify which SRTM raster tile to return
(SRTM data come in large raster tiles that cover the globe). As usual, look
at the getData function documentation for a description of the arguments.
To estimate the PEF’s centroid coordinate, we averaged the forest inventory
plots’ latitude and longitude then assigned the result to pef.centroid (for fun,
I used the apply function to do the same task when creating mu.centroid
earlier in the chapter). Second, srtm is in a longitude latitude geographic
CRS (same as our other PEF data). Finally, the image shows SRTM elevation
192 8 Spatial Data Visualization and Analysis

along the coast of Maine, the PEF plots are those tiny specks of black in the
northwest quadrant, and the white region of the image is the Atlantic Ocean.
Okay, this is a start, but it would be good to crop the SRTM image to the
PEF’s extent. This is done using raster’s crop function. This function can
use many different kinds of spatial objects in the second argument to calculate
the extent at which to crop the object in the first argument. Here, I set mu as
the second argument and save the resulting SRTM subset over the larger tile
(the srtm object).

srtm <- crop(srtm, mu)

image(srtm)
plot(mu, add = TRUE)
44.860
y

44.850
44.840

-68.640 -68.630 -68.620 -68.610

The crop is in effect doing a spatial subsetting of the raster data. We’ll return
to the srtm data and explore different kinds of subsetting in the subsequent
sections.

8.7.2 Logical, Index, and Name Subsetting

As promised, we can subset spatial objects using the [] operator and a logical,
index, or name vector. The key is that sp objects behave like data frames, see
Section 4.5. A logical or index vector to the left of the comma in [,] accesses
8.7 Subsetting Spatial Data 193

points, lines, polygons, or pixels. Similarly, a logical, index, or name vector to

the right of the comma accesses variables.
For example, say we want to map forest inventory plots with more than 10,000
stems per hectare, stems_ha (the min() was added below to double check that
the subset worked correctly.

min(plots$stems_ha)

[1] 119

plots.10k <- plots[plots$stems_ha > 10000,]

min(plots.10k$stems_ha)

[1] 10008
You can also add new variables to the spatial objects.

plots$diameter_in <- plots$diameter_cm/2.54

head(plots)

mu_id plot biomass_mg_ha stems_ha diameter_cm

1 U10 11 96.35 5453 4.8
2 U10 13 115.70 2629 6.9
3 U10 21 121.58 3385 6.1
4 U10 22 71.97 7742 3.1
5 U10 23 122.26 7980 4.7
6 U10 24 85.85 10047 1.6
basal_area_m2_ha diameter_in
1 22.0 1.8898
2 23.2 2.7165
3 23.0 2.4016
4 16.1 1.2205
5 29.2 1.8504
6 19.1 0.6299

8.7.3 Spatial Subsetting and Overlay

A spatial overlay retrieves the indexes or variables from object A using the
location of object B. With some spatial objects this operation can be done using
the [] operator. For example, say we want to select and map all management
194 8 Spatial Data Visualization and Analysis

units in mu, i.e., A, that contain plots with more than 10,000 stems per ha, i.e.,
B.

mu.10k <- mu[plots.10k,]## A[B,]

mu.10k.f <- fortify(mu.10k, region="mu_id")

ggmap(basemap) +
geom_polygon(data=mu.10k.f, aes(x = long, y = lat, group=group), fill="transparent", s
geom_point(data=as.data.frame(plots.10k), aes(x = easting, y = northing), color="white

44.87

44.86
lat

44.85

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

More generally, however, the over function offers consistent overlay operations
for sp objects and can return either indexes or variables from object A given
locations from object B, i.e., over(B, A) or, equivalently, B%over%A. The code
below duplicates the result from the preceding example using over.

mu.10k <- mu[mu$mu_id %in% unique(over(plots.10k, mu)$mu_id),]

Yes, this requires more code but over provides a more flexible and general
purpose function for overlays on the variety of sp objects. Let’s unpack this
one-liner into its five steps.
8.7 Subsetting Spatial Data 195

i <- over(plots.10k, mu)

ii <- i$mu_id
iii <- unique(ii)
iv <- mu$mu_id %in% iii
v <- mu[iv,]

i. The over function returns variables for mu’s polygons that coincide
with the nrow(plots.10k@data) points in plots.10k. No points
fall outside the polygons and the polygons do not overlap, so i should
be a data frame with nrow(plots.10k@data) rows. If polygons did
overlap and a point fell within the overlap region, then variables for
the coinciding polygons are returned.
ii. Select the unique mu identifier mu_id (this step is actually not nec-
essary here because mu only has one variable).
iii. Because some management units contain multiple plots there will
be repeat values of mu_id in ii, so apply the unique function to get
rid of duplicates.
iv. Use the %in% operator to create a logical vector that identifies which
polygons should be in the final map.
v. Subset mu using the logical vector created in iv.

Now let’s do something similar using the srtm elevation raster. Say we want
to map elevation along trails, winter roads, and gravel roads across the PEF.
We could subset srtm using the roads SpatialLinesDataFrame; however,
mapping the resulting pixel values along the road segments using ggmap
requires a bit more massaging. So, to simplify things for this example, roads is
first coerced into a SpatialPointsDataFrame called roads.pts that is used
to extract spatially coinciding srtm pixel values which themselves are coerced
from raster’s RasterLayer to sp’s SpatialPixelsDataFrame called srtm.sp
so that we can use the over function. We also choose a different basemap just
for fun.

hikes <- roads[roads$type %in% c("Trail", "Winter", "Gravel"),]

hikes.pts <- as(hikes, "SpatialPointsDataFrame")

srtm.sp <- as(srtm, "SpatialPixelsDataFrame")

hikes.pts$srtm <- over(hikes.pts, srtm.sp)

basemap <- get_map(location=mu.centroid, zoom=14, maptype = "terrain", source = "google")

color.vals <- srtm@data@values[1:length(hikes.pts)]

196 8 Spatial Data Visualization and Analysis

ggmap(basemap) +
geom_point(data=as.data.frame(hikes.pts),
aes(x = coords.x1, y = coords.x2, color = color.vals)) +
scale_color_gradient(low="green", high="red") +
labs(color = "Hiking trail elevation\n(m above sea level)",
xlab="Longitude", ylab="Latitude")

Warning: Removed 483 rows containing missing values

(geom_point).

44.87

44.86
Hiking trail elevation
(m above sea level)
80
70
lat

60
44.85
50
40
30

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

In the call to geom_point above, coords.x1 coords.x2 are the default names
given to longitude and latitude, respectively, when sp coerces hikes to
hikes.pts. These points represent the vertices along line segments. I cre-
ate the vector color.vals that contains the values from srtm that I use in
the map argument color.
Overlay operations involving lines and polygons over polygons require the
rgeos package which provides an interface to the Geometry Engine - Open
Source28 (GEOS) C++ library for topology operations on geometries. We’ll
leave it to you to explore these capabilities.
28 https://trac.osgeo.org/geos/
8.7 Subsetting Spatial Data 197

8.7.4 Spatial Aggregation

We have seen aggregation operations before when using dplyr’s summarize

function. The summarize function is particularly powerful when combined
with group_by(), which can apply a function specified in summarize() to a
variable partitioned using a grouping variable. The aggregate function in sp
works in a similar manner, except groups are delineated by the spatial extent
of a thematic variable. In fact, the work we did to create mu.bio using dplyr
functions can be accomplished with aggregate(). Using aggregate() will,
however, require a slightly different approach for joining the derived average
biomass_mg_ha to the fortified mu. This is because the aggregate function
will apply the user specified function to all variables in the input object, which,
in our case, results in an NA for the linking variable mu_id as demonstrated
below.

mu.ag <- aggregate(plots[,c("mu_id","biomass_mg_ha")], by=mu, FUN=mean)

head(mu.ag@data, n=2)

mu_id biomass_mg_ha
0 <NA> 49.86
1 <NA> 112.17
With mu_id rendered useless, we do not have a variable that uniquely identifies
each polygon for use in fortify’s region argument; hence no way to subse-
quently join the unfortified and fortified versions of mu.bio.ag. Here’s the
work around. If the region is not specified, fortify() uses an internal unique
polygon ID that is part of the sp data object and accessed via row.names()29
So, the trick is to add this unique polygon ID to the aggregate() output prior
to calling fortify() as demonstrated below.

mu.ag$id <- row.names(mu.ag)

mu.ag.f <- fortify(mu.ag)

Regions defined for each Polygons

mu.ag.f <- left_join(mu.ag.f, mu.ag@data)

29 With other data, there is a chance the row names differ from the unique poly-

gon IDs. Therefore a more reliable approach to getting a unique ID is to use

sapply(slot(mu.ag, 'polygons'), function(x) slot(x, 'ID')), but replace mu.ag with
your SpatialPolygonsDataFrame. Also, this approach will work with other sp objects in the
right column of Table 8.1.
198 8 Spatial Data Visualization and Analysis

Joining, by = "id"

ggmap(basemap) +
geom_polygon(data=mu.ag.f, aes(x = long, y = lat,
group=group, fill=biomass_mg_ha),
size=0.2, color="orange") +
scale_fill_gradient(low="white", high="darkblue",
na.value="transparent") +
labs(fill="Biomass (mg/ha)")

44.87

44.86
Biomass (mg/ha)

150
lat

100
44.85

50

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

The aggregate() function will work with all sp objects. For example let’s
map the variance of pixel values in srtm.sp by management unit. Notice that
aggregate() is happy to take a user-specified function for FUN.

mu.srtm <- aggregate(srtm.sp, by=mu,

FUN=function(x){sqrt(var(x))})

mu.srtm$id <- row.names(mu.srtm)

mu.srtm.f <- fortify(mu.srtm)

Regions defined for each Polygons

8.8 Where to go from here 199

mu.srtm.f <- left_join(mu.srtm.f, mu.srtm@data)

Joining, by = "id"

ggmap(basemap) +
geom_polygon(data=mu.srtm.f, aes(x = long, y = lat, group=group,
fill=srtm_23_04),
size=0.2, color="orange") +
scale_fill_gradient(low="green", high="red") +
labs(fill = "Elevation standard deviation\n(m above sea level)",
xlab="Longitude", ylab="Latitude")

44.87

44.86 Elevation standard deviation

(m above sea level)
6
lat

4
44.85
2

44.84

-68.65 -68.64 -68.63 -68.62 -68.61 -68.60

lon

8.8 Where to go from here

This chapter just scratches the surface of R’s spatial data manipulation and
visualization capabilities. The basic ideas we presented here should allow you
to take a deeper look into sp, rgdal, rgeos, ggmap, leaflet, and a myriad of
other excellent user-contributed R spatial data packages. A good place to start
is with Edzer Pebesma’s excellent vignette on the use of the map overlay and
200 8 Spatial Data Visualization and Analysis

spatial aggregation, available here30 , as well as Applied Spatial Data Analysis

with R by Bivand et al. (2013).

8.9 Exercises

Exercise Spatial Data Learning objectives: practice loading and reprojecting

spatial data; analyze spatial data; create leaflet maps to convey analysis results;
interpret analysis results.

30 https://cran.r-project.org/web/packages/sp/vignettes/over.pdf
9
Shiny: Interactive Web Apps in R

Shiny1 is a framework that turns R code and figures into interactive web
applications. Let’s start out by looking at a built-in example Shiny app.

library(shiny)
runExample("01_hello")

In the bottom panel of the resulting Shiny app (Figure 9.1), we can see the
R script that is essential to running any Shiny app: app.R. Take a minute to
explore how the app works and how the script code is structured. The first part
of the script (ui <-) defines the app’s user interface (UI) using directives that
partition the resulting web page and placement of input and output elements.
The second part of the script defines a function called server with arguments
input and output. This function provides the logic required to run the app
(in this case, to draw the histogram of the Old Faithful Geyser Data). Third,
ui and server are passed to a function called shinyApp which creates the
application.2
All of the example code and data for the remainder of this chapter can be
accessed using the following code.

library(downloader)

download("https://www.finley-lab.com/files/data/shiny_chapter_code.zip",
destfile="./shiny_chapter_code.zip", mode="wb")

unzip("shiny_chapter_code.zip", exdir = ".")

1 https://shiny.rstudio.com/
2 Since 2014, Shiny has supported single-file applications (one file called app.R that

contains UI and server components), but in other resources, you may see two separate source
files, server.R and ui.R, that correspond to those two components. We will use the updated
one-file system here, but keep in mind that older resources you find on the internet using
the Shiny package may employ the two-file approach. Ultimately, the code inside these files
is almost identical to that within the single app.R file. See https://shiny.rstudio.com/ar
ticles/app-formats.html for more information.

201
202 9 Shiny: Interactive Web Apps in R

FIGURE 9.1: Hello Shiny example app

9.1 Running a Simple Shiny App

Shiny apps allow users to interact with a data set. Let’s say we have access to
a dummy data set name_list.csv. Begin by taking a quick look at what this
CSV file contains.

namesDF <- read.csv("https://www.finley-lab.com/files/data/name_list.csv",

stringsAsFactors = FALSE)
str(namesDF)

'data.frame': 200 obs. of 4 variables:

$ First: chr "Sarah" "Boris" "Jessica" "Diane" ...
$ Last : chr "Poole" "Dowd" "Wilkins" "Murray" ...
9.1 Running a Simple Shiny App 203

$ Age : int 15 52 12 58 56 4 14 0 28 27 ...

$ Sex : chr "F" "M" "F" "F" ...
Given the namesDF data, our goal over the next few sections is to develop
a Shiny application that allows users to explore age distributions for males
and females. We begin by defining a very basic Shiny application script upon
which we can build the desired functionality. The script file that defines this
basic application is called app.R and is provided in the code block below called
“version 1”. Follow the subsequent steps to create and run this Shiny app
(note these steps are already done for you in the “ShinyPractice” directory in
“shiny_chapter_code.zip”, but you might want to do it yourself):

1. Create a new directory called “ShinyPractice”.

2. In the “ShinyPractice” directory, create a blank R script called
app.R.
3. Copy the code in “app.R version 1” into app.R.
4. Run the Shiny app from RStudio. There are two ways to do this:
1) use the RStudio button (See Figure 9.2) or; 2) type the function
runApp() in the RStudio console.

# app.R version 1

library(shiny)

names.df <- read.csv("https://www.finley-lab.com/files/data/name_list.csv")

# Define UI
ui <- fluidPage(

titlePanel("Random Names Analysis"),

sidebarLayout(

sidebarPanel("our inputs will be here"),

mainPanel("our output will appear here")

)
)

# Define server logic

server <- function(input, output) {}

# Create Shiny app

shinyApp(ui = ui, server = server)
204 9 Shiny: Interactive Web Apps in R

FIGURE 9.2: Running a Shiny app locally

If everything compiles correctly, you should get a window that looks like
Figure 9.3. There are a few things to notice. First, we read in the CSV file
at the beginning of the app.R script so that these data can be accessed in
subsquent development steps (although we don’t use it yet in this initial
application). Second, our Shiny app is not interactive yet because there is no
way for a user to input any data—the page is view-only. Third, the function
function(input,output) {} in app.R does not include any code, reflecting
that we are not yet using any user inputs or creating any outputs.

FIGURE 9.3: Your first Shiny app

9.2 Adding User Input 205

9.2 Adding User Input

Shiny package widgets are used to collect user provided inputs (visit the
Shiny Widget Gallery3 to get a sense of the different kinds of input dialogues).
In the subsequent development we’ll use slider bars, dropdown menus, and
check boxes. All widget functions, at minimum, include an inputId and label
argument. The character string assigned to the inputId argument will be the
variable name used to access the value(s) selected or provided by the user.
This variable name is used in the application development and not seen by the
user. The character string assigned to the label argument appears in the UI
above the given input widget and serves to provide a user-friendly description
of the desired input.
Building on the “version 1” code, we add a dropdown menu to the script’s UI
portion using the selectInput function to the “version 2” code below. Update
app.R and rerun the application. What is our inputId for the dropdown menu?
What is our label? Can you see where the label is displayed in the app when
you rerun it?
Look at the structure of the choices argument in selectInput. It’s a vector
of options in the form "name" = "value" where the name is what the user
sees, and the value is what is returned to the server logic function of the
app.R file.

# app.R version 2

library(shiny)

names.df <- read.csv("https://www.finley-lab.com/files/data/name_list.csv")

# Define UI
ui <- fluidPage(
titlePanel("Random Names Age Analysis"),
sidebarLayout(
sidebarPanel(
# Dropdown selection for Male/Female
selectInput(inputId = "sexInput", label = "Sex:",
choices = c("Female" = "F",
"Male" = "M",
"Both" = "B"))
),

3 https://shiny.rstudio.com/gallery/widget-gallery.html
206 9 Shiny: Interactive Web Apps in R

mainPanel("our output will appear here")

)
)

# Define server logic

server <- function(input, output) {}

# Create Shiny app

shinyApp(ui = ui, server = server)

9.3 Adding Output

Now that we’ve included the user input dialog, let’s make the application
truly interactive by changing the output depending on user input. This is
accomplished by modifying the server logic portion of the script. Our goal is
to plot an age distribution histogram in the main panel given the sex selected
by the user.

9.3.1 Interactive Server Logic

Server logic is defined by two arguments: input and output. These objects
are both list-like, so they can contain multiple other objects. We already know
from the user input part of the app.R script that we have an input component
called sexInput, which can be accessed in the reactive portion of the server
logic by calling input$sexInput (notice the use of the $ to access the input
value associated with sexInput). In “version 3” of the application, we use
the information held in input$sexInput to subset names.df then create a
histogram of names.df$Age. This histogram graphic is included as an element
in the output object and ultimately made available in the UI portion of the
script by referencing its name histogram.
Reactive portions of our app.R script’s server logic are inside the server func-
tion. We create reactive outputs using Shiny’s functions like renderPlot.4
Obviously, renderPlot renders the contents of the function into a plot ob-
ject; this is then stored in the output list wherever we assign it (in this case,
output$histogram). Notice the contents of the renderPlot function are con-
4 Every reactive output function’s name in Shiny is of the form render*.
9.3 Adding Output 207

tained not only by regular parentheses, but also by curly brackets (just one
more piece of syntax to keep track of).

# app.R version 3

library(shiny)

names.df <- read.csv("https://www.finley-lab.com/files/data/name_list.csv")

# Define UI
ui <- fluidPage(
titlePanel("Random Names Age Analysis"),
sidebarLayout(
sidebarPanel(
# Dropdown selection for Male/Female
selectInput(inputId = "sexInput", label = "Sex:",
choices = c("Female" = "F",
"Male" = "M",
"Both" = "B"))
),
mainPanel("our output will appear here")
)
)

# Define server logic

server <- function(input, output) {

output$histogram <- renderPlot({

if(input$sexInput != "B"){
subset.names.df <- subset(names.df, Sex == input$sexInput)
} else {
subset.names.df <- names.df
}

ages <- subset.names.df$Age

# draw the histogram with the specified 20 bins

hist(ages, col = 'darkgray', border = 'white')
})

}
208 9 Shiny: Interactive Web Apps in R

# Create Shiny app

shinyApp(ui = ui, server = server)

Update your app.R server logic function to match the code above. Rerun the
application. Note the appearance of the app doesn’t change because we have
not updated the UI portion with the resulting histogram.

9.3.2 Interactive User Interface

Now we update the UI part of app.R to make the app interactive. In the
“version 4” code below, the plotOutput("histogram") function in ui accesses
the histogram component of the output list and plots it in the main panel.
Copy the code below and rerun the application. You have now created your
first Shiny app!

# app.R version 4

library(shiny)

names.df <- read.csv("https://www.finley-lab.com/files/data/name_list.csv")

# Define UI
ui <- fluidPage(
titlePanel("Random Names Age Analysis"),
sidebarLayout(
sidebarPanel(
# Dropdown selection for Male/Female
selectInput(inputId = "sexInput", label = "Sex:",
choices = c("Female" = "F",
"Male" = "M",
"Both" = "B"))
),
mainPanel(plotOutput("histogram"))
)
)

# Define server logic

server <- function(input, output) {

output$histogram <- renderPlot({

if(input$sexInput != "B"){
9.4 More Advanced Shiny App: Michigan Campgrounds 209

subset.names.df <- subset(names.df, Sex == input$sexInput)

} else {
subset.names.df <- names.df
}

ages <- subset.names.df$Age

# draw the histogram with the specified 20 bins

hist(ages, col = 'darkgray', border = 'white')
})

# Create Shiny app

shinyApp(ui = ui, server = server)

9.4 More Advanced Shiny App: Michigan Campgrounds

The Michigan Department of Natural Resources (DNR) has made substantial

investments in open-sourcing its data via the Michigan DNR’s open data5
initiative. We’ll use some of these data to motivate our next Shiny application.
Specifically we’ll work with the DNR State Park Campgrounds data available
here6 as a CSV file.

u <- "https://www.finley-lab.com/files/data/Michigan_State_Park_Campgrounds.csv"
sites <- read.csv(u, stringsAsFactors = F)
str(sites)

'data.frame': 151 obs. of 19 variables:

$ X : num -84.4 -84.4 -83.8 -83.8 -83.8 ...
$ Y : num 42.9 42.9 42.5 42.5 42.5 ...
$ OBJECTID : int 10330 10331 5986 5987 5988 5989 5990 5991 5992 5993 ...
$ Type : chr "Campground" "Campground" "Campground" "Campground" ...
$ Detail_Type: chr "State Park" "State Park" "State Park" "State Park" ...
$ DISTRICT : chr "Rose Lake" "Rose Lake" "Rose Lake" "Rose Lake" ...
$ COUNTY : chr "Clinton" "Clinton" "Livingston" "Livingston" ...
$ FACILITY : chr "Sleepy Hollow State Park" "Sleepy Hollow State Park" "Brighton Recre
5 http://gis-midnr.opendata.arcgis.com/datasets/
6 https://www.finley-lab.com/files/data/Michigan_State_Park_Campgrounds.csv
 210 9 Shiny: Interactive Web Apps in R

$ Camp_type : chr "Modern" "Organizational" "Modern" "Organizational" ...

$ TOTAL_SITE : int 203 1 144 10 25 25 219 75 2 297 ...
$ ADA_SITES : int 0 0 0 0 0 0 0 3 0 13 ...
$ name : chr "Sleepy Hollow State Park " "Sleepy Hollow State Park " "Brighton R
$ Ownership : chr "State" "State" "State" "State" ...
$ Management : chr "DNR" "DNR" "DNR" "DNR" ...
$ Surface : chr "Dirt" "Dirt" "Dirt" "Dirt" ...
$ Use_ : chr "Recreation" "Recreation" "Recreation" "Recreation" ...
$ Condition : chr "Unknown" "Unknown" "Unknown" "Unknown" ...
$ Lat : num 42.9 42.9 42.5 42.5 42.5 ...
$ Long : num -84.4 -84.4 -83.8 -83.8 -83.8 ...
We see that Michigan has 151 state park campgrounds, and our data frame
contains 19 variables. Let’s create a Shiny app UI in which the user selects
desired campground specifications, and the app displays the list of resulting
campgrounds and their location on a map. The complete Shiny app.R is
provided in the “CampsitesMI” directory in “shiny_chapter_code.zip.” Start
out by running the application to see how it works. Examine the example
code and how it relates to the application that you are running. The following
sections detail each of the app’s components.

9.4.1 Michigan Campgrounds UI

First, let’s look at the structure of the page. Similar to our first application,
we again use a fluidPage layout with title panel and sidebar. The sidebar
contains a sidebar panel and a main panel. Our sidebar panel has three user
input widgets:
• sliderInput: Allows user to specify a range of campsites desired in their
campground. Since the maximum number of campsites in any Michigan state
park campground is 411, 420 was chosen as the maximum.
• selectInput: Allows user to select what type of campsites they want. To
get the entire list of camp types, we used the data frame, sites, loaded at
the beginning of the script.
• checkboxInput: Allows the user to see only campgrounds with ADA sites
available.
Table 9.1 provides a list of the various input and output elements. Take your
time and track how the app defines then uses each input and output.

9.4.2 Michigan Campgrounds Server Logic

In creating our server variable, we have two functions that fill in our output
elements:
 9.5 Adding Leaflet to Shiny 211

TABLE 9.1: Interactive elements

Element ID Description Function to Create

input$rangeNum desired range of campsite quantity sliderInput
input$type desired campsite type selectInput
input$ada desired ADA site availability checkboxInput
output$plot1 map with campground markers renderPlot
output$text1 HTML-formatted list of campgrounds renderText

• renderText: Creates a character string of HTML code to print the bulleted

list of available sites.
• renderPlot: Creates a ggplot2 map of Michigan with campground markers.
Note that both of these functions contain identical subsetting of the sites
data frame into the smaller sites1 data frame (see below). As you can see
from the code, we use the three inputs from the application to subset the data:
rangeNum from the slider widget, type from the dropdown menu, and ada
from the checkbox.
This repeated code can be avoided using Shiny’s reactive expressions. These
expressions will update in value whenever the widget input values change. See
here7 for more information. The use of reactive expressions is beyond the scope
of this chapter, but it is an important concept to be familiar with if you plan
to regularly create Shiny applications.

sites1 <- subset(sites,

TOTAL_SITE >= input$rangeNum[1] &
TOTAL_SITE <= input$rangeNum[2] &
Camp_type == input$type &
if(input$ada){ ADA_SITES > 0 } else {ADA_SITES >= 0}
)

9.5 Adding Leaflet to Shiny

Leaflet8 can be easily incorporated in Shiny apps (See Chapter 8 of this text to
learn more about spatial data). In the Michigan Campgrounds example code,
we used plotOutput and renderPlot to put a plot widget in our Shiny app.
7 https://shiny.rstudio.com/tutorial/written-tutorial/lesson6/
8 https://rstudio.github.io/leaflet
212 9 Shiny: Interactive Web Apps in R

Similarly, in this code, we will use leafletOutput and renderLeaflet to add

Leaflet widgets to our app. Run the Shiny app in the CampsitesMI_Leaflet
directory. What are the differences between the plot widget in this app and
the plot widget in the previous app (using ggplot2)?
Our code inside renderLeaflet is displayed below. As a reminder from our
previous use of Leaflet, we can use the magrittr package’s pipe operator, %>%,
to add properties to our Leaflet plot. In the addCircleMarkers function, you
can see that we used the mapply function to apply HTML code to each of the
markers. The HTML code simply prints the site name and number of sites
within each marker label. The else statement serves the purpsose that if our
subset, sites1, is empty, we render a map centered on Lansing, MI.
}

# renderLeaflet function from app.R

output$plot1 <- renderLeaflet({

# create a subset of campsites based on inputs
sites1 <- subset(sites,
TOTAL_SITE >= input$rangeNum[1] &
TOTAL_SITE <= input$rangeNum[2] &
Camp_type == input$type &
if(input$ada){ ADA_SITES > 0 } else {ADA_SITES >= 0})

if(nrow(sites1) > 0){

leaflet(sites1) %>% addTiles() %>%
addCircleMarkers(lng = ~Long, lat = ~Lat,
radius = 5,
color = "red",
label = mapply(function(x,y) {
HTML(sprintf('<em>%s</em><br>%s site(s)',
htmlEscape(x),
htmlEscape(y)))},
sites1$FACILITY, sites1$TOTAL_SITE
)
)
} else {
leaflet() %>% addTiles() %>%
setView( -84.5555, 42.7325, zoom = 7)
}

})
 9.7 Why use Shiny? 213

9.6 Why use Shiny?

In this chapter, we learned what a Shiny app is, what it’s components are,
how to run the app, and how to incorporate Leaflet plots. We can also host
these apps online on our own server or the shinyapps.io9 server, which can be
accessed directly from RStudio. Hosting our apps on a server allows anyone
with internet access to interact with our widgets.
The shinyapps.io server allows free hosting of apps within the monthly limits
of 5 applications and 25 active hours. Paid plans are also available. A user
guide to deploying your application on shinyapps.io is available here10 . Set up
a free account and impress your friends by sending them links to your Shiny
apps! In addition to hosting your applications, here are a couple more things
to discover with Shiny:
• Dive deeper11 into development by working through the RStudio Shiny
tutorial series.
• Incorporate12 CSS stylesheets to make your apps fancier.

9.7 Exercises
Exercise Shiny Learning objectives: practice updating ggplot2 plot aesthet-
ics; modify Shiny HTML output; add an interactive Shiny element.

9 https://www.shinyapps.io/
10 http://docs.rstudio.com/shinyapps.io/index.html
11 https://shiny.rstudio.com/tutorial/
12 https://shiny.rstudio.com/articles/css.html
10
Classification

Classification problems are quite common. For example a spam filter is asked
to classify incoming messages into spam or non-spam, based on factors such as
the sender’s address, the subject of the message, the contents of the message,
and so on. As another example, a doctor diagnosing a patient into one of four
possible diagnoses based on symptoms, blood tests, and medical history is
another form of classification. Or a bank may want to determine (prior to
giving a loan) whether an applicant for a loan will default on the loan, again
based on several variables such as income, financial history, etc.
Classification methods are both extremely useful and an active area of research
in statistics. In this chapter we will learn about two common, and somewhat
different, classification methods, logistic regression and k nearest neighbors. A
very good introduction to classification and many other “Statistical Learning”
methods is James et al. (2014). The abbreviated treatment in this chapter
draws from James et al. (2014).

10.1 Logistic Regression

Logistic regression is widely used to relate a categorical response variable to
one or more (continuous or categorical) predictors. Initially we will consider the
simplest case where the response Y has only two possible values (we’ll assume
the values are 0 and 1) and where there is only one continuous predictor X. We
would like to predict the value of Y based on the value of the predictor X. Let
p(X) = P (Y = 1|X).1 . We will model p(X) with the logistic function, which
takes values between 0 and 1, which is of course appropriate for modeling a
probability:

eβ0 +β1 X
p(X) = .
1 + eβ0 +β1 X
1 For those not familiar with probability notation, the right side of this equation reads as

“The probability that Y = 1 given we know the value of X”)

215
216 10 Classification

A graph of this function when β0 = 0 and β1 = 1 shows the characteristic

shape.

library(ggplot2)
logistic <- function(x){exp(x)/(1 + exp(x))}
ggplot(data.frame(x = c(-6, 6)), aes(x)) +
stat_function(fun = logistic)

1.00

0.75

0.50
y

0.25

0.00

-6 -3 0 3 6
x

To make this more concrete, we will consider a data set on a population of

women who were over 21 years of age, lived near Phoenix, Arizona, and were
of Pima Indian heritage. The data included diabetes status (yes or no) and
seven possible predictor variables such as age, number of pregnancies, body
mass index, etc. The original data are from the National Institute of Diabetes
and Digestive and Kidney Diseases. These data were cleaned and are available
in the MASS package in R. In the package the data frame Pima.tr contains
200 randomly selected cases from the full data set, which we will use to find a
classifier. The data frame Pima.te contains the remaining 332 cases, which we
will use to test the classifier on new data.2

library(MASS)
head(Pima.tr)

2 This practice of randomly separating data into training data and testing data is a

common practice for many classification methods.

10.1 Logistic Regression 217

npreg glu bp skin bmi ped age type

1 5 86 68 28 30.2 0.364 24 No
2 7 195 70 33 25.1 0.163 55 Yes
3 5 77 82 41 35.8 0.156 35 No
4 0 165 76 43 47.9 0.259 26 No
5 0 107 60 25 26.4 0.133 23 No
6 5 97 76 27 35.6 0.378 52 Yes
It will be more convenient to code the presence or absence of diabetes by 1
and 0, so we first create another column in the data frame with this coding:

Pima.tr$diabetes <- rep(0, dim(Pima.tr)[1])

Pima.tr$diabetes[Pima.tr$type == "Yes"] <- 1
head(Pima.tr)

npreg glu bp skin bmi ped age type diabetes

1 5 86 68 28 30.2 0.364 24 No 0
2 7 195 70 33 25.1 0.163 55 Yes 1
3 5 77 82 41 35.8 0.156 35 No 0
4 0 165 76 43 47.9 0.259 26 No 0
5 0 107 60 25 26.4 0.133 23 No 0
6 5 97 76 27 35.6 0.378 52 Yes 1
We will begin with glu as a predictor, and of course type as the response. So
we want to find the values β0 and β1 which provide the best fit for the model

eβ0 +β1 (glu)

P (yes|glu) =
1 + eβ0 +β1 (glu)

Usually maximum likelihood methods are used to fit the model.

In R we will use the function glm to fit logistic regression models. The glm
function fits a wide variety of models. To specify the logistic regression model
we specify family = binomial as an argument to glm. We also must specify
the predictor and response variables via the model formula, which in our case
will be diabetes ~ glu to indicate that diabetes (i.e., type recoded) is the
response and glu is the predictor; also we specify the data frame, which in
our case is Pima.tr.

diabetes.lr1 <- glm(diabetes ~ glu, data = Pima.tr, family = binomial)

diabetes.lr1

Call: glm(formula = diabetes ~ glu, family = binomial, data = Pima.tr)

Coefficients:
(Intercept) glu
218 10 Classification

-5.5036 0.0378

Degrees of Freedom: 199 Total (i.e. Null); 198 Residual

Null Deviance: 256
Residual Deviance: 207 AIC: 211

summary(diabetes.lr1)

Call:
glm(formula = diabetes ~ glu, family = binomial, data = Pima.tr)

Deviance Residuals:
Min 1Q Median 3Q Max
-1.971 -0.779 -0.529 0.849 2.263

Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -5.50364 0.83608 -6.58 0.000000000046
glu 0.03778 0.00628 6.02 0.000000001756

(Intercept) ***
glu ***
---
Signif. codes:
0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

Null deviance: 256.41 on 199 degrees of freedom

Residual deviance: 207.37 on 198 degrees of freedom
AIC: 211.4

Number of Fisher Scoring iterations: 4

beta0.lr.1 <- coef(diabetes.lr1)[1]

beta1.lr.1 <- coef(diabetes.lr1)[2]
beta0.lr.1

(Intercept)
-5.504

beta1.lr.1

glu
10.1 Logistic Regression 219

0.03778
The coefficients β0 and β1 are approximately -5.504 and 0.038, respectively. So
for example we can estimate the probability that a woman in this population
whose glucose level is 150 would be diabetic as

exp(-5.504 + 0.038*150)/(1 + exp(-5.504 + 0.038*150))

[1] 0.5488
We can plot the fitted probabilities along with the data “by hand”.

diabetes.logistic.1 <- function(x){

exp(beta0.lr.1+ beta1.lr.1*x)/(1 + exp(beta0.lr.1+
beta1.lr.1*x))
}
ggplot(Pima.tr, aes(x = glu, y = diabetes)) +
stat_function(fun = diabetes.logistic.1) + geom_point()

1.00

0.75
diabetes

0.50

0.25

0.00

50 100 150 200

glu

The ggplot2 package also provides a way to do this more directly, using
stat_smooth.

ggplot(Pima.tr, aes(x = glu, y = diabetes)) +

geom_point() +
220 10 Classification

stat_smooth(method = "glm",
method.args = list(family = "binomial"), se = FALSE)

`geom_smooth()` using formula 'y ~ x'

1.00

0.75
diabetes

0.50

0.25

0.00

50 100 150 200

glu

From these graphics we can see that although glucose level and diabetes are
related, there are many women with high glucose levels who are not diabetic,
and many with low glucose levels who are diabetic, so likely adding other
predictors to the model will help.
Next let’s see how the model does in predicting diabetes status in the data we
did not use for fitting the model. We will predict diabetes for anyone whose
glucose level leads to a model-based probability greater than 1/2. First we
use the predict function to compute the probabilities, and then use these to
make predictions.

head(Pima.te)

npreg glu bp skin bmi ped age type

1 6 148 72 35 33.6 0.627 50 Yes
2 1 85 66 29 26.6 0.351 31 No
3 1 89 66 23 28.1 0.167 21 No
4 3 78 50 32 31.0 0.248 26 Yes
5 2 197 70 45 30.5 0.158 53 Yes
10.1 Logistic Regression 221

6 5 166 72 19 25.8 0.587 51 Yes

diabetes.probs.1 <- predict(diabetes.lr1, Pima.te, type = "response")

head(diabetes.probs.1)

1 2 3 4 5 6
0.52207 0.09179 0.10519 0.07199 0.87433 0.68319
The predict function (with type = "response" specified) provides p(x) =
P (Y = 1|X = x) for all the x values in a data frame. In this case we specified
the data frame Pima.te since we want to know how the model does in predicting
diabetes in a new population, i.e., in a population that wasn’t used to “train”
the model.

diabetes.predict.1 <- rep("No", dim(Pima.te)[1])

diabetes.predict.1[diabetes.probs.1 > 0.5] <- "Yes"
head(diabetes.predict.1)

[1] "Yes" "No" "No" "No" "Yes" "Yes"

table(diabetes.predict.1, Pima.te$type)

diabetes.predict.1 No Yes
No 206 58
Yes 17 51

length(diabetes.predict.1[diabetes.predict.1 == Pima.te$type])/
dim(Pima.te)[1]

[1] 0.7741
The table (sometimes called a confusion matrix) has the predictions of the
model in the rows, so for example we see that the model predicts that 206+58 =
264 of the women will not be diabetic, and that 17 + 51 = 68 of the women will
be diabetic. More interesting of course are the cells themselves. For example,
of the 206 + 17 = 223 women who are not diabetic in Pima.te, the model
correctly classifies 206, and misclassifies 17. A classifier that predicted perfectly
for the test data would have zeros off the diagonal.

10.1.1 Adding Predictors

If we have p predictors X = (X1 , . . . , Xp ), the logistic model becomes

222 10 Classification

eβ0 +β1 X+···+βp Xp

p(X) = p(X) = .
1 + eβ0 +β1 X+···+βp Xp

Although there is a lot more notation to keep track of, the basic ideas are the
same as they were for the one predictor model. We will next see how adding
bmi, the body mass index, affects predictions of diabetes.

diabetes.lr2 <- glm(diabetes ~ glu + bmi,

data = Pima.tr, family = binomial)
diabetes.lr2

Call: glm(formula = diabetes ~ glu + bmi, family = binomial, data = Pima.tr)

Coefficients:
(Intercept) glu bmi
-8.2161 0.0357 0.0900

Degrees of Freedom: 199 Total (i.e. Null); 197 Residual

Null Deviance: 256
Residual Deviance: 198 AIC: 204

summary(diabetes.lr2)

Call:
glm(formula = diabetes ~ glu + bmi, family = binomial, data = Pima.tr)

Deviance Residuals:
Min 1Q Median 3Q Max
-2.058 -0.757 -0.431 0.801 2.249

Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -8.21611 1.34697 -6.10 0.0000000011
glu 0.03572 0.00631 5.66 0.0000000152
bmi 0.09002 0.03127 2.88 0.004

(Intercept) ***
glu ***
bmi **
---
Signif. codes:
0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

10.1 Logistic Regression 223

Null deviance: 256.41 on 199 degrees of freedom

Residual deviance: 198.47 on 197 degrees of freedom
AIC: 204.5

Number of Fisher Scoring iterations: 4

Now we look at predictions from this model.

diabetes.probs.2 <- predict(diabetes.lr2, Pima.te, type = "response")

head(diabetes.probs.2)

1 2 3 4 5 6
0.52359 0.05810 0.07530 0.06662 0.82713 0.50879

diabetes.predict.2 <- rep("No", dim(Pima.te)[1])

diabetes.predict.2[diabetes.probs.2 > 0.5] <- "Yes"
head(diabetes.predict.2)

[1] "Yes" "No" "No" "No" "Yes" "Yes"

table(diabetes.predict.2, Pima.te$type)

diabetes.predict.2 No Yes
No 204 54
Yes 19 55

length(diabetes.predict.2[diabetes.predict.2 == Pima.te$type])/
dim(Pima.te)[1]

[1] 0.7801
Adding bmi as a predictor did not improve the predictions by very much!
Let x1 and x2 represent glucose and bmi levels. We classify a subject as
“diabetic” if the fitted p(X) is greater than 0.5, i.e., if the fitted probability of
diabetes is greater than the fitted probability of not diabetes. The boundary
for our decision is where these two fitted probabilities are equal, i.e., where

P (Y = 1|(x1 , x2 )) P (Y = 1|(x1 , x2 ))
= = 1.
P (Y = 0|(x1 , x2 )) 1 − P (Y = 1|(x1 , x2 ))

Writing these out in terms of the logistic regression model, taking logarithms,
and performing some algebra leads to the following (linear!) decision boundary:
224 10 Classification

β0 β1
x2 = − − x1 .
β2 β2

The diabetes training data along with the decision boundary are plotted below.

lr.int <- -coef(diabetes.lr2)[1]/coef(diabetes.lr2)[3]

lr.slope <- -coef(diabetes.lr2)[2]/coef(diabetes.lr2)[3]
ggplot(Pima.tr, aes(x = glu, y = bmi)) +
geom_point(aes(color = type)) +
geom_abline(intercept = lr.int, slope = lr.slope)

40

type
bmi

No
Yes
30

20

50 100 150 200

glu

10.1.2 More than Two Classes

Logistic regression methods also are applicable to classification contexts where

there are more than two classes. Consider for example Fisher’s iris data.

data(iris)
head(iris)

Sepal.Length Sepal.Width Petal.Length Petal.Width

1 5.1 3.5 1.4 0.2
2 4.9 3.0 1.4 0.2
10.1 Logistic Regression 225

3 4.7 3.2 1.3 0.2

4 4.6 3.1 1.5 0.2
5 5.0 3.6 1.4 0.2
6 5.4 3.9 1.7 0.4
Species
1 setosa
2 setosa
3 setosa
4 setosa
5 setosa
6 setosa

str(iris)

'data.frame': 150 obs. of 5 variables:

$ Sepal.Length: num 5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
$ Sepal.Width : num 3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
$ Petal.Length: num 1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
$ Petal.Width : num 0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
$ Species : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...

ggplot(data = iris, aes(x = Sepal.Length, y = Sepal.Width)) +

geom_point(aes(color = Species))

4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor
3.0
virginica

2.5

2.0

5 6 7 8
Sepal.Length
226 10 Classification

ggplot(data = iris, aes(x = Petal.Length, y = Petal.Width)) +

geom_point(aes(color = Species))

2.5

2.0

Species
1.5
Petal.Width

setosa
versicolor
virginica
1.0

0.5

0.0
2 4 6
Petal.Length

Here the potential predictors are sepal width, sepal length, petal width, and
petal length. The goal is to find a classifier that will yield the correct species.
From the scatter plots it should be pretty clear that a model with petal length
and petal width as predictors would classify the data well. Although in a sense
this is too easy, these data do a good job of illustrating logistic regression with
more than two classes.
Before doing that we randomly choose 75 of the 150 rows of the data frame to
be the training sample, with the other 75 being the test sample.

set.seed(321)
selected <- sample(1:150, replace = FALSE, size = 75)
iris.train <- iris[selected,]
iris.test <- iris[-selected,]

There are several packages which implement logistic regression for data with
more than two classes. We will use the VGAM package. The function vglm within
VGAM implements logistic regression (and much more).
10.1 Logistic Regression 227

library(VGAM)
iris.lr <- vglm(Species ~ Petal.Width + Petal.Length,
data = iris.train, family = multinomial)
summary(iris.lr)

Call:
vglm(formula = Species ~ Petal.Width + Petal.Length, family = multinomial,
data = iris.train)

Pearson residuals:
Min 1Q
log(mu[,1]/mu[,3]) -0.0000397 0.0000000175
log(mu[,2]/mu[,3]) -1.5279112 -0.0020907887
Median 3Q Max
log(mu[,1]/mu[,3]) 0.0000000257 0.0000000395 0.0000796
log(mu[,2]/mu[,3]) 0.0000000972 0.0101234167 1.6385058

Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept):1 104.97 8735.20 NA NA
(Intercept):2 52.71 27.05 NA NA
Petal.Width:1 -41.55 21914.63 NA NA
Petal.Width:2 -9.18 4.49 -2.04 0.041 *
Petal.Length:1 -18.93 8993.53 NA NA
Petal.Length:2 -7.62 4.90 -1.55 0.120
---
Signif. codes:
0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Names of linear predictors: log(mu[,1]/mu[,3]),

log(mu[,2]/mu[,3])

Residual deviance: 14.65 on 144 degrees of freedom

Log-likelihood: -7.324 on 144 degrees of freedom

Number of Fisher scoring iterations: 21

Warning: Hauck-Donner effect detected in the following estimate(s):

'(Intercept):1', '(Intercept):2', 'Petal.Width:1', 'Petal.Length:1'

Reference group is level 3 of the response

228 10 Classification

Notice that the family is specified as multinomial rather than binomial since
we have more than two classes. When run with these data, the vglm function
returns several (about 20) warnings. These occur mainly because the classes
are so easily separated, and are suppressed above.
Next we compute the probabilities for the test data.

iris.probs <- predict(iris.lr, iris.test[,c(3,4)], type="response")

head(iris.probs)

setosa versicolor
2 1 0.00000000000009874
5 1 0.00000000000009874
6 1 0.00000000190561922
7 1 0.00000000000251434
10 1 0.00000000000001202
11 1 0.00000000000030599
virginica
2 0.00000000000000000000000000000034016
5 0.00000000000000000000000000000034016
6 0.00000000000000000000000040468850399
7 0.00000000000000000000000000002168557
10 0.00000000000000000000000000000003543
11 0.00000000000000000000000000000225863
At least for the first six cases, one probability is close to one and the other
two are close to zero, reflecting the fact that this is an easy classification
problem. Next we extract the actual predictions. For these, we want to choose
the highest probability in each row. The which.max function makes this easy.
Before applying this to the fitted probabilities, we illustrate its use. Take notice
that which.max() returns the position of the highest probability, and not the
actualy highest probability itself.

which.max(c(2,3,1,5,8,3))

[1] 5

which.max(c(2,20,4,5,9,1,0))

[1] 2

class.predictions <- apply(iris.probs, 1, which.max)

head(class.predictions)
10.2 Nearest Neighbor Methods 229

2 5 6 7 10 11
1 1 1 1 1 1

class.predictions[class.predictions == 1] <- levels(iris$Species)[1]

class.predictions[class.predictions == 2] <- levels(iris$Species)[2]
class.predictions[class.predictions == 3] <- levels(iris$Species)[3]
head(class.predictions)

2 5 6 7 10 11
"setosa" "setosa" "setosa" "setosa" "setosa" "setosa"
Next we create the confusion matrix.

table(class.predictions, iris.test$Species)

class.predictions setosa versicolor virginica

setosa 25 0 0
versicolor 0 22 1
virginica 0 0 27

10.2 Nearest Neighbor Methods

Nearest neighbor methods provide a rather different way to construct classifiers,
and have strengths (minimal assumptions, flexible decision boundaries) and
weaknesses (computational burden, lack of interpretability) compared to logistic
regression models.
In principle the idea is simple. Recall that the training set will have both x
and y values known, while the test set will have only x values known. Begin
by choosing a positive integer k which will specify the number of neighbors
to use in classification. To classify a point x in the training set, find the k
closest x values in the training set, and choose the class which has the highest
representation among the k points. The algorithm is called kNN for “k Nearest
Neighbors”.
For example, suppose that k = 10 and the 10 nearest neighbors to a training
x have classes 1, 1, 3, 2, 3, 3, 3, 2, 3, 2. Since there are five 3s, three 2s, and two
1s, the training point is assigned to class 3. Suppose that for another x the 10
nearest neighbors have classes 1, 1, 1, 2, 3, 1, 3, 3, 3, 2. In this case there are four
1s and four 3s, so there is a tie for the lead. The nearest neighbor algorithm
then will choose between 1 and 3 at random.
230 10 Classification

Although in principle kNN is simple, some issues arise. First, how should k be
chosen? There is not an easy answer, but it can help to think of the extreme
values for k.
The largest possible k is the number of observations in the training set.
For example suppose that the training set has 10 observations, with classes
1, 1, 1, 2, 2, 2, 3, 3, 3, 3. Then for any point in the test set, the k = 10 nearest
neighbors will include ALL of the points in the training set, and hence every
point in the test set will be classified in class 3. This classifier has low (zero)
variance, but probably has high bias.
The smallest possible k is 1. In this case, each point in the test set is put in the
same class as its nearest neighbor in the training set. This may lead to a very
non-smooth and high variance classifier, but the bias will tend to be small.
A second issue that is relatively easy to deal with concerns the scales on which
the x values are measured. If for example one x variable has a range from 2 to
4, while another has a range from 2000 to 4000, the distance between the test
and training data will be dominated by the second variable. The solution that
is typically used is to standardize all the variables (rescale them so that their
mean is 0 and their standard deviation is 1).
These and other issues are discussed in the literature on kNN, but we won’t
pursue them further.
There are at least three R packages which implement kNN, including class,
kknn, and RWeka. We will use class below.
An example from Hastie et al. (2009) will be used to give a better sense of the
role of k in the kNN algorithm. The example uses simulated data and shows
the decision boundaries for kNN with k = 15 and k = 1.3 . Although the R
code used to draw the displays is given below, focus mainly on the graphics
produced, and what they tell us about kNN.
First the data are read into R and graphed. The predictors x1 and x2, while not
standardized, have very similar standard deviations, so we will not standardize
these data before applying kNN.

u.knn <- "https://www.finley-lab.com/files/data/knnExample.csv"

knnExample <- read.csv(u.knn, header=TRUE)
str(knnExample)

'data.frame': 200 obs. of 3 variables:

$ x1 : num 2.5261 0.367 0.7682 0.6934 -0.0198 ...
$ x2 : num 0.3211 0.0315 0.7175 0.7772 0.8673 ...
3 The graphs below use code adapted from http://stackoverflow.com/questions/

31234621/variation-on-how-to-plot-decision-boundary-of-a-k-nearest-neighbor-
classifier-f
10.2 Nearest Neighbor Methods 231

$ class: int 0 0 0 0 0 0 0 0 0 0 ...

ggplot(data = knnExample, aes(x = x1, y = x2)) +

geom_point(aes(color = as.factor(class))) +
theme_bw()

1
as.factor(class)
0
x2

1
0

-1

-2

-2 0 2 4
x1

Next a large set of test data is created using the expand.grid function, which
creates a data frame with all possible combinations of the arguments. First
a simple example to illustrate the function, then the actual creation of the
training set. The test set covers the range of the x1 and x2 values in the
training set.

expand.grid(x = c(1,2), y = c(5,3.4,2))

x y
1 1 5.0
2 2 5.0
3 1 3.4
4 2 3.4
5 1 2.0
6 2 2.0
232 10 Classification

min(knnExample$x1)

[1] -2.521

max(knnExample$x1)

[1] 4.171

min(knnExample$x2)

[1] -2

max(knnExample$x2)

[1] 2.856

x.test <- expand.grid(x1 = seq(-2.6, 4.2, by=0.1), x2 = seq(-2, 2.9, by=0.1))

Next the kNN with k = 15 is fit. Notice that the first argument gives the x
values in the training set, the next argument gives the x values in the test set,
the third argument gives the y values (labels) from the training set. The fourth
argument gives k, and the fifth argument asks for the function to return the
probabilities of membership (that is, the proportion of the nearest neighbors
which were in the majority class) as well as the class assignments.

library(class)
Example_knn <- knn(knnExample[,c(1,2)], x.test, knnExample[,3], k = 15, prob = TRUE)
prob <- attr(Example_knn, "prob")
head(prob)

[1] 0.6667 0.6667 0.6667 0.7333 0.7333 0.7333

Next the graphs are created. This is somewhat complex, since we want to plot
the test data colored by the class they were assigned to by the kNN classifier
as background, the training data (using a different symbol), and the decision
boundary.

library(dplyr)
df1 <- mutate(x.test, prob = prob, class = 0, prob_cls = ifelse(Example_knn == class, 1,
str(df1)
10.2 Nearest Neighbor Methods 233

'data.frame': 3450 obs. of 5 variables:

$ x1 : num -2.6 -2.5 -2.4 -2.3 -2.2 -2.1 -2 -1.9 -1.8 -1.7 ...
$ x2 : num -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 ...
$ prob : num 0.667 0.667 0.667 0.733 0.733 ...
$ class : num 0 0 0 0 0 0 0 0 0 0 ...
$ prob_cls: num 1 1 1 1 1 1 1 1 1 1 ...
- attr(*, "out.attrs")=List of 2
..$ dim : Named int [1:2] 69 50
.. ..- attr(*, "names")= chr [1:2] "x1" "x2"
..$ dimnames:List of 2
.. ..$ x1: chr [1:69] "x1=-2.6" "x1=-2.5" "x1=-2.4" "x1=-2.3" ...
.. ..$ x2: chr [1:50] "x2=-2.0" "x2=-1.9" "x2=-1.8" "x2=-1.7" ...

df2 <- mutate(x.test, prob = prob, class = 1, prob_cls = ifelse(Example_knn == class, 1,

bigdf <- bind_rows(df1, df2)

names(knnExample)

[1] "x1" "x2" "class"

ggplot(bigdf) +
geom_point(aes(x=x1, y =x2, col=class), data = mutate(x.test, class = Example_knn), siz
geom_point(aes(x = x1, y = x2, col = as.factor(class)), size = 4, shape = 1, data = knn
geom_contour(aes(x = x1, y = x2, z = prob_cls, group = as.factor(class), color = as.fac

1
class
0
x2

1
0

-1

-2

-2 0 2 4
x1
234 10 Classification

Next we graph the decision boundary of kNN with k = 1.

Example_knn <- knn(knnExample[,c(1,2)], x.test, knnExample[,3], k = 1, prob = TRUE)

prob <- attr(Example_knn, "prob")
head(prob)

[1] 1 1 1 1 1 1

df1 <- mutate(x.test, prob = prob, class = 0, prob_cls = ifelse(Example_knn == class, 1,

str(df1)

'data.frame': 3450 obs. of 5 variables:

$ x1 : num -2.6 -2.5 -2.4 -2.3 -2.2 -2.1 -2 -1.9 -1.8 -1.7 ...
$ x2 : num -2 -2 -2 -2 -2 -2 -2 -2 -2 -2 ...
$ prob : num 1 1 1 1 1 1 1 1 1 1 ...
$ class : num 0 0 0 0 0 0 0 0 0 0 ...
$ prob_cls: num 1 1 1 1 1 1 1 1 1 1 ...
- attr(*, "out.attrs")=List of 2
..$ dim : Named int [1:2] 69 50
.. ..- attr(*, "names")= chr [1:2] "x1" "x2"
..$ dimnames:List of 2
.. ..$ x1: chr [1:69] "x1=-2.6" "x1=-2.5" "x1=-2.4" "x1=-2.3" ...
.. ..$ x2: chr [1:50] "x2=-2.0" "x2=-1.9" "x2=-1.8" "x2=-1.7" ...

df2 <- mutate(x.test, prob = prob, class = 1, prob_cls = ifelse(Example_knn == class, 1,

bigdf <- bind_rows(df1, df2)

ggplot(bigdf) + geom_point(aes(x = x1, y = x2, col = class), data = mutate(x.test, class =

10.2 Nearest Neighbor Methods 235

1
class
0
x2

1
0

-1

-2

-2 0 2 4
x1

10.2.1 kNN and the Diabetes Data

Next kNN is applied to the diabetes data. We will use the same predictors,
glu and bmi that were used in the logistic regression example. Since the scales
of the predictor variables are substantially different, they are standardized
first. The value k = 15 is chosen for kNN.

Pima.tr[,1:7] <- scale(Pima.tr[,1:7], center = TRUE, scale = TRUE)

Pima.te[,1:7] <- scale(Pima.te[,1:7], center = TRUE, scale = TRUE)
knn_Pima <- knn(Pima.tr[,c(2,5)], Pima.te[,c(2,5)], Pima.tr[,8], k = 15, prob=TRUE)
table(knn_Pima, Pima.te[,8])

knn_Pima No Yes
No 206 55
Yes 17 54
At least in terms of the confusion matrix, kNN with k = 15 performed about
as well as logistic regression for these data.
236 10 Classification

10.2.2 Practice Problem

Produce a figure that displays how the number of false positives produced
from the kNN classifier for the diabetes data set changes for all integer values
of k from 1 to 40. Use this graph to justify whether or not k = 15 was a valid
choice for the number of neighbors.

10.2.3 kNN and the iris Data

Now kNN is used to classify the iris data. As before we use petal length
and width as predictors. The scales of the two predictors are not particularly
different, so we won’t standardize the predictors. Unsurprisingly kNN does
well in classifying the test set for a wide variety of k values.

sd(iris.train$Petal.Width)

[1] 0.7316

sd(iris.train$Petal.Length)

[1] 1.705

head(iris.train)

Sepal.Length Sepal.Width Petal.Length Petal.Width

77 6.8 2.8 4.8 1.4
88 6.3 2.3 4.4 1.3
58 4.9 2.4 3.3 1.0
96 5.7 3.0 4.2 1.2
126 7.2 3.2 6.0 1.8
17 5.4 3.9 1.3 0.4
Species
77 versicolor
88 versicolor
58 versicolor
96 versicolor
126 virginica
17 setosa

knn_iris <- knn(iris.train[,c(3,4)], iris.test[,c(3,4)], iris.train[,5], k=1, prob=TRUE)

table(knn_iris, iris.test[,5])
10.3 Exercises 237

knn_iris setosa versicolor virginica

setosa 25 0 0
versicolor 0 22 1
virginica 0 0 27

knn_iris <- knn(iris.train[,c(3,4)], iris.test[,c(3,4)], iris.train[,5], k=3, prob=TRUE)

table(knn_iris, iris.test[,5])

knn_iris setosa versicolor virginica

setosa 25 0 0
versicolor 0 22 1
virginica 0 0 27

knn_iris <- knn(iris.train[,c(3,4)], iris.test[,c(3,4)], iris.train[,5], k=15, prob=TRUE)

table(knn_iris, iris.test[,5])

knn_iris setosa versicolor virginica

setosa 25 0 0
versicolor 0 22 1
virginica 0 0 27

10.3 Exercises
Exercise Classification Learning objectives: explore the logistic regression
classification method; apply the kNN classification method; create confusion
matrices to compare classification methods; plot classified data.
11
Text Data

Many applications require the ability to manipulate and process text data.
For example, an email spam filter takes as its input various features of email
such as the sender, words in the subject, words in the body, the number and
types of attachments, and so on. The filter then tries to build a classifier which
can correctly classify a message as spam or not spam (aka ham). As another
example, some works of literature, such as some of Shakespeare’s plays or
some of the Federalist papers, have disputed authorship. By analyzing word
use across many documents, researchers try to determine the author of the
disputed work.
Working with text data requires functions that will, for example, concatenate
and split text strings, modify strings (e.g., converting to lower-case or removing
vowels), count the number of characters in a string, and so on. In addition to
being useful in such contexts, string manipulation is helpful more generally in
R—for example, to effectively construct titles for graphics.
As with most tasks, there are a variety of ways to accomplish these text
processing tasks in R. The base R package has functions which work with and
modify text strings. Another useful package which approaches these tasks in a
slightly different way is stringr. As with graphics, we will focus mainly on
one package to avoid confusion. In this case we will focus on the base R string
processing functions, but will emphasize that stringr is also worth knowing.
The application to analyzing Moby Dick below comes from the book {Text
Analysis with R for Students of Literature by Matthew L. Jockers.

11.1 Reading Text Data into R

Often text data will not be in a rectangular format that is suitable for reading
into a data frame. For example, an email used to help train a spam filter, or
literary texts used to help determine authorship of a novel are certainly not
of this form. Often when working with text data we want to read the whole
text object into a single R vector. In this case either the scan function or

239
240 11 Text Data

the readLines function are useful. The readLines function is typically more
efficient, but scan is much more flexible.
As an example, consider the following email message and a plain text version of
the novel Moby Dick by Herman Melville, the beginning of which is displayed
subsequently.
From safety33o@l11.newnamedns.com Fri Aug 23 11:03:37 2002
Return-Path: <safety33o@l11.newnamedns.com>
Delivered-To: zzzz@localhost.example.com
Received: from localhost (localhost [127.0.0.1])
by phobos.labs.example.com (Postfix) with ESMTP id 5AC994415F
for <zzzz@localhost>; Fri, 23 Aug 2002 06:02:59 -0400 (EDT)
Received: from mail.webnote.net [193.120.211.219]
by localhost with POP3 (fetchmail-5.9.0)
for zzzz@localhost (single-drop); Fri, 23 Aug 2002 11:02:59 +0100 (IST)
Received: from l11.newnamedns.com ([64.25.38.81])
by webnote.net (8.9.3/8.9.3) with ESMTP id KAA09379
for <zzzz@example.com>; Fri, 23 Aug 2002 10:18:03 +0100
From: safety33o@l11.newnamedns.com
Date: Fri, 23 Aug 2002 02:16:25 -0400
Message-Id: <200208230616.g7N6GOR28438@l11.newnamedns.com>
To: kxzzzzgxlrah@l11.newnamedns.com
Reply-To: safety33o@l11.newnamedns.com
Subject: ADV: Lowest life insurance rates available!
moode

Lowest rates available for term life insurance! Take a moment

and fill out our online form
to see the low rate you qualify for.
Save up to 70% from regular rates! Smokers accepted!
http://www.newnamedns.com/termlife/

Representing quality nationwide carriers. Act now!

The Project Gutenberg EBook of Moby Dick; or The Whale, by Herman Melville

This eBook is for the use of anyone anywhere at no cost and with
almost no restrictions whatsoever. You may copy it, give it away or
re-use it under the terms of the Project Gutenberg License included
with this eBook or online at www.gutenberg.org

Title: Moby Dick; or The Whale

Author: Herman Melville

11.1 Reading Text Data into R 241

Last Updated: January 3, 2009

Posting Date: December 25, 2008 [EBook #2701]
Release Date: June, 2001

Language: English

*** START OF THIS PROJECT GUTENBERG EBOOK MOBY DICK; OR THE WHALE ***

Produced by Daniel Lazarus and Jonesey

MOBY DICK; OR THE WHALE

By Herman Melville
The email message is available at https://www.finley-lab.com/files/da
ta/email1.txt while the novel is available at https://www.finley-lab.co
m/files/data/mobydick.txt. We will read these into R using scan.
First, we read in the email message. The scan function has several possible
arguments. For now the important arguments are the file to be read (the
argument is named file), the type of data in the file (the argument is named
what), and how the fields in the file are separated (the argument is named
sep). To illustrate the sep argument, the file will be read into R once with
sep = "" indicating that the separator is whitespace, and once with sep =
"\n" indicating that the separator is the newline character, i.e., each field in
the file is a line.

u.email <- "https://www.finley-lab.com/files/data/email1.txt"

email1 <- scan(u.email, what = "character", sep = "")
length(email1)

[1] 133

email1[1:10]

[1] "From"
242 11 Text Data

[2] "safety33o@l11.newnamedns.com"
[3] "Fri"
[4] "Aug"
[5] "23"
[6] "11:03:37"
[7] "2002"
[8] "Return-Path:"
[9] "<safety33o@l11.newnamedns.com>"
[10] "Delivered-To:"

email1 <- scan(u.email, what = "character", sep = "\n")

length(email1)

[1] 26

email1[1:10]

[1] "From safety33o@l11.newnamedns.com Fri Aug 23 11:03:37 2002"

[2] "Return-Path: <safety33o@l11.newnamedns.com>"
[3] "Delivered-To: zzzz@localhost.example.com"
[4] "Received: from localhost (localhost [127.0.0.1])"
[5] "\tby phobos.labs.example.com (Postfix) with ESMTP id 5AC994415F"
[6] "\tfor <zzzz@localhost>; Fri, 23 Aug 2002 06:02:59 -0400 (EDT)"
[7] "Received: from mail.webnote.net [193.120.211.219]"
[8] "\tby localhost with POP3 (fetchmail-5.9.0)"
[9] "\tfor zzzz@localhost (single-drop); Fri, 23 Aug 2002 11:02:59 +0100 (IST)"
[10] "Received: from l11.newnamedns.com ([64.25.38.81])"
Note that when sep = "" was specified, every time whitespace was encountered
R moved to a new element of the vector email1, and this vector ultimately
contained 133 elements. When sep = "\n" was specified, all the text before
a newline was put into one element of the vector, which ended up with 26
elements.
The scan function is quite flexible. In fact, read.table uses scan to actually
read in the data. Read the help file for scan if more information is desired.
Next Moby Dick is read in line by line.

u.moby <- "https://www.finley-lab.com/files/data/mobydick.txt"

moby_dick <- scan(u.moby, what = "character", sep = "\n")
moby_dick[1:25]

[1] "The Project Gutenberg EBook of Moby Dick; or The Whale, by Herman Melville"
[2] "This eBook is for the use of anyone anywhere at no cost and with"
11.1 Reading Text Data into R 243

[3] "almost no restrictions whatsoever. You may copy it, give it away or"
[4] "re-use it under the terms of the Project Gutenberg License included"
[5] "with this eBook or online at www.gutenberg.org"
[6] "Title: Moby Dick; or The Whale"
[7] "Author: Herman Melville"
[8] "Last Updated: January 3, 2009"
[9] "Posting Date: December 25, 2008 [EBook #2701]"
[10] "Release Date: June, 2001"
[11] "Language: English"
[12] "*** START OF THIS PROJECT GUTENBERG EBOOK MOBY DICK; OR THE WHALE ***"
[13] "Produced by Daniel Lazarus and Jonesey"
[14] "MOBY DICK; OR THE WHALE"
[15] "By Herman Melville"
[16] "Original Transcriber's Notes:"
[17] "This text is a combination of etexts, one from the now-defunct ERIS"
[18] "project at Virginia Tech and one from Project Gutenberg's archives. The"
[19] "proofreaders of this version are indebted to The University of Adelaide"
[20] "Library for preserving the Virginia Tech version. The resulting etext"
[21] "was compared with a public domain hard copy version of the text."
[22] "In chapters 24, 89, and 90, we substituted a capital L for the symbol"
[23] "for the British pound, a unit of currency."
[24] "ETYMOLOGY."
[25] "(Supplied by a Late Consumptive Usher to a Grammar School)"
You will notice that the scan function ignored blank lines in the file. If it is
important to preserve blank lines, the argument blank.lines.skip = FALSE
can be supplied to scan.
The file containing the novel contains some introductory and closing text that
is not part of the original novel. If we are interested in Melville’s writing, we
should remove this text. By inspection we can discover that the novel’s text
begins at position 408 and ends at position 18576.

moby_dick <- moby_dick[408:18576]

length(moby_dick)

[1] 18169

moby_dick[1:4]

[1] "CHAPTER 1. Loomings."

[2] "Call me Ishmael. Some years ago--never mind how long precisely--having"
[3] "little or no money in my purse, and nothing particular to interest me on"
[4] "shore, I thought I would sail about a little and see the watery part of"
244 11 Text Data

moby_dick[18165:18169]

[1] "they glided by as if with padlocks on their mouths; the savage sea-hawks"
[2] "sailed with sheathed beaks. On the second day, a sail drew near, nearer,"
[3] "and picked me up at last. It was the devious-cruising Rachel, that in"
[4] "her retracing search after her missing children, only found another"
[5] "orphan."

11.2 The paste Function

The paste function concatenates vectors after (if necessary) converting the
vectors to character.

paste("Homer Simpson", "is", "Bart Simpson's", "father")

[1] "Homer Simpson is Bart Simpson's father"

n <- 10
paste("The value of n is", n)

[1] "The value of n is 10"

paste(c("pig", "dog"), 3)

[1] "pig 3" "dog 3"

By default the paste function separates the input vectors with a space. But
other separators can be specified.

paste("mail", "google", "com", sep=".")

[1] "mail.google.com"

paste("and", "or", sep = "/")

[1] "and/or"
11.2 The paste Function 245

paste(c("dog", "cat", "horse", "human", "elephant"), "food")

[1] "dog food" "cat food" "horse food"

[4] "human food" "elephant food"
Sometimes we want to take a character vector with n elements and create a
character vector with only one element, which contains all n character strings.
Setting the collapse argument to something other than the default NULL tells
R we want to do this, and allows specification of the separator in the collapsed
vector.

paste(c("one", "two", "three", "four", "five"),

c("six", "seven", "eight", "nine", "ten"))

[1] "one six" "two seven" "three eight"

[4] "four nine" "five ten"

paste(c("one", "two", "three", "four", "five"),

c("six", "seven", "eight", "nine", "ten"), collapse = ".")

[1] "one six.two seven.three eight.four nine.five ten"

paste(c("one", "two", "three", "four", "five"),

c("six", "seven", "eight", "nine", "ten"), collapse = "&&")

[1] "one six&&two seven&&three eight&&four nine&&five ten"

paste(c("one", "two", "three", "four", "five"),

c("six", "seven", "eight", "nine", "ten"), collapse = " ")

[1] "one six two seven three eight four nine five ten"
In the example above by default paste created a vector with five elements,
each containing one input string from the first input vector and one from
the second vector, pasted together. When a non NULL argument was specified
for collapse, the vector created had one element, with the pasted strings
separated by that argument.1
1 There is a somewhat subtle difference among the examples. If all the arguments are

length one vectors, then paste by default returns a length one vector. But if one or more of
the arguments have length greater than one, the default behavior of paste is to return a
vector of length greater than one. The collapse argument changes this behavior.
246 11 Text Data

Also don’t forget that R “recycles” values from vectors if two or more different
length vectors are provided as input.

paste(c("a", "b"), 1:10, sep = "")

[1] "a1" "b2" "a3" "b4" "a5" "b6" "a7" "b8"

[9] "a9" "b10"

paste(c("a", "b"), 1:9, sep = "")

[1] "a1" "b2" "a3" "b4" "a5" "b6" "a7" "b8" "a9"
Next, consider writing a function which simulates repeatedly tossing a coin n
times, counting the number of HEADS out of the n tosses. For the first five
repetitions of n tosses, the function will print out the number of HEADS (for
example if there are 7 HEADS in the n = 10 tosses the function prints “The
number of HEADS out of 10 tosses is 7.” The function returns a histogram of
the number of HEADS, with a title stating “Number of HEADS in ?? tosses”
where ?? is replaced by the number of tosses. The paste function will help
greatly.

coin_toss <- function(n=10, iter = 500){

require(ggplot2)
df <- data.frame(numheads = numeric(iter))
for(i in 1:iter) {
df$numheads[i] <- rbinom(1, n, 0.5)
if(i <= 5) {
print(paste("The number of HEADS out of", n, "tosses is", df$numheads[i]))}
}
ggplot(data = df, aes(x = numheads)) +
geom_histogram(binwidth = 1) +
ggtitle(paste("Number of HEADS in", n, "tosses"))
}
coin_toss()

[1] "The number of HEADS out of 10 tosses is 5"

[1] "The number of HEADS out of 10 tosses is 7"
[1] "The number of HEADS out of 10 tosses is 8"
[1] "The number of HEADS out of 10 tosses is 4"
[1] "The number of HEADS out of 10 tosses is 4"
11.2 The paste Function 247

Number of HEADS in 10 tosses

100
count

50

0.0 2.5 5.0 7.5 10.0

numheads

coin_toss(n = 25, iter=1000)

[1] "The number of HEADS out of 25 tosses is 13"

[1] "The number of HEADS out of 25 tosses is 10"
[1] "The number of HEADS out of 25 tosses is 16"
[1] "The number of HEADS out of 25 tosses is 13"
[1] "The number of HEADS out of 25 tosses is 13"
248 11 Text Data

Number of HEADS in 25 tosses

150

100
count

50

5 10 15 20
numheads

Let’s now return to the object moby_dick that contains the text of the novel.
If we want to analyze word choice, word frequency, etc., it would be helpful to
form a vector in which each element is a word from the novel. One way to do
this is to first paste the current version of the moby_dick variable into a new
version which is one long vector with the lines pasted together. To illustrate,
we will first do this with a much smaller object that shares the structure of
moby_dick.

small_novel <- c("First line", "Second somewhat longer line",

"third line.")
small_novel

[1] "First line"

[2] "Second somewhat longer line"
[3] "third line."

small_novel <- paste(small_novel, collapse=" ")

length(small_novel)

[1] 1

small_novel

[1] "First line Second somewhat longer line third line."

11.3 More String Processing Functions 249

Now we do the same with the actual novel.

moby_dick <- paste(moby_dick, collapse = " ")

length(moby_dick)

[1] 1
At this point moby_dick contains a single very long character string. Next we
will separate this string into separate words and clean up the resulting vector
a bit.

11.3 More String Processing Functions

Common string processing tasks include changing case between upper and
lower, extracting and/or replacing substrings of a string, trimming a string to
a specified width, counting the number of characters in a string, etc.

11.3.1 tolower and toupper

R contains functions tolower and toupper which very simply change the case
of all characters in a string.

x <- "aBCdefG12#"
y <- x
tolower(x)

[1] "abcdefg12#"

toupper(y)

[1] "ABCDEFG12#"
If we are interested in frequencies of words in Moby Dick, converting all the text
to the same case makes sense, so for example the word “the” at the beginning
of a sentence is not counted differently than the same word in the middle of a
sentence.

moby_dick <- tolower(moby_dick)

250 11 Text Data

11.3.2 nchar and strsplit

The function nchar counts the number of characters in a string or strings.

nchar("dog")

[1] 3

nchar(c("dog", "cat", "horse", "elephant"))

[1] 3 3 5 8

nchar(c("dog", "cat", "horse", "elephant", NA, "goat"))

[1] 3 3 5 8 NA 4

nchar(c("dog", "cat", "horse", "elephant", NA, "goat"), keepNA = FALSE)

[1] 3 3 5 8 2 4

nchar(moby_dick)

[1] 1190309
By default nchar returns NA for a missing value. If you want nchar to return
2 for a NA value, you can set keepNA = TRUE.2
The function strsplit splits the elements of a character vector. The function
returns a list, and often the unlist function is useful to convert the list into
an atomic vector.

strsplit(c("mail.msu.edu", "mail.google.com", "www.amazon.com"),

split = ".", fixed = TRUE)

[[1]]
[1] "mail" "msu" "edu"

[[2]]
[1] "mail" "google" "com"

2 It may be reasonable if the purpose of counting characters is to find out how much space

to allocate for printing a vector of strings where the NA string will be printed.
11.3 More String Processing Functions 251

[[3]]
[1] "www" "amazon" "com"

unlist(strsplit(c("mail.msu.edu", "mail.google.com", "www.amazon.com"),

split = ".", fixed = TRUE))

[1] "mail" "msu" "edu" "mail" "google"

[6] "com" "www" "amazon" "com"

unlist(strsplit(c("dog", "cat", "pig", "horse"),

split = "o", fixed = TRUE))

[1] "d" "g" "cat" "pig" "h" "rse"

Setting the argument fixed to TRUE tells R to match the value of split
exactly when performing the split. The function can be much more powerful if
the value of split is a regular expression, which can for example ask for splits
at any vowels, etc. We will not go in depth on Regular Expressions here, but
we will make some use of regular expressions on a case-by-case basis prior to
that. Regular expressions are very powerful, so if this chapter interests you,
we suggest researching regular expressions on your own (as always there are
plenty of free resources online).

unlist(strsplit(c("dog", "cat", "pig", "horse", "rabbit"),

split = "[aeiou]"))

[1] "d" "g" "c" "t" "p" "g" "h" "rs" "r" "bb"
[11] "t"
The regular expression [aeiou] represents any of the letters a, e, i, o, u. In
general a string of characters enclosed in square brackets indicates any one
character in the string.

unlist(strsplit(c("dog", "cat", "pig", "horse", "rabbit"),

split = "[aorb]"))

[1] "d" "g" "c" "t" "pig" "h" "" "se"

[9] "" "" "" "" "it"
The regular expression [aorb] represents any of the letters a, o, r, b.

unlist(strsplit(c("a1c2b", "bbb2bc3f"), split = "[1-9]"))

[1] "a" "c" "b" "bbb" "bc" "f"

252 11 Text Data

The regular expression [1-9] represents any of the numbers 1, 2, 3, 4, 5, 6, 7,

8, 9.

unlist(strsplit(c("aBc1fGh", "1TyzaaG"), split = "[ˆa-z]"))

[1] "a" "c" "f" "h" "" "" "yzaa"

The regular expression [a-z] represents any lower case letter. The caret ˆ
in front of a-z indicates “match any character except those in the following
string” which in this case indicates “match any character that is NOT a lower
case letter”.
Recall that the moby_dick vector now contains one long character string which
includes the entire text of the novel, and that we would like to split it into
separate words. We now know how to do this using strsplit and a regular
expression. First a smaller example.

unlist(strsplit(c("the rain", "in Spain stays mainly", "in", "the plain"),

split = "[ˆ0-9A-Za-z]"))

[1] "the" "rain" "in" "Spain" "stays"

[6] "mainly" "in" "the" "plain"

unlist(strsplit(c("the rain", "in Spain stays mainly", "in", "the plain"),

split =" ", fixed = TRUE))

[1] "the" "rain" "in" "Spain" "stays"

[6] "mainly" "in" "the" "plain"
Look at the regular expression. The caret says “match anything but” and
then 0=9A-Za-z says “any digit, any lower-case letter, and any upper-case
letter.” So the whole expression (including the fact that it is the value of the
argument split) says “match anything but any digit, any lower-case letter, or
any upper-case letter”.
Now we apply this to moby_dick.

moby_dick <- unlist(strsplit(moby_dick, split = "[ˆ0-9A-Za-z]"))

Let’s see a bit of what we have.

moby_dick[1:50]

[1] "chapter" "1" ""

11.3 More String Processing Functions 253

[4] "loomings" "" "call"

[7] "me" "ishmael" ""
[10] "some" "years" "ago"
[13] "" "never" "mind"
[16] "how" "long" "precisely"
[19] "" "having" "little"
[22] "or" "no" "money"
[25] "in" "my" "purse"
[28] "" "and" "nothing"
[31] "particular" "to" "interest"
[34] "me" "on" "shore"
[37] "" "i" "thought"
[40] "i" "would" "sail"
[43] "about" "a" "little"
[46] "and" "see" "the"
[49] "watery" "part"
There is a small problem: Some of the “words” are blank. The following small
example indicates why this happened.

unlist(strsplit(c("the rain", "in Spain stays mainly", "in", "the plain"), split = "[ˆ0

[1] "the" "rain" "in" "Spain" ""

[6] "" "" "stays" "mainly" "in"
[11] "the" "plain"
It is not too hard to remove the blank words.

length(moby_dick)

[1] 253993

not.blanks <- which(moby_dick != "")

moby_dick <- moby_dick[not.blanks]
length(moby_dick)

[1] 214889

moby_dick[1:50]

[1] "chapter" "1" "loomings"

[4] "call" "me" "ishmael"
[7] "some" "years" "ago"
[10] "never" "mind" "how"
254 11 Text Data

[13] "long" "precisely" "having"

[16] "little" "or" "no"
[19] "money" "in" "my"
[22] "purse" "and" "nothing"
[25] "particular" "to" "interest"
[28] "me" "on" "shore"
[31] "i" "thought" "i"
[34] "would" "sail" "about"
[37] "a" "little" "and"
[40] "see" "the" "watery"
[43] "part" "of" "the"
[46] "world" "it" "is"
[49] "a" "way"
(In this example it would have been more efficient to replace

moby_dick <- unlist(strsplit(moby_dick, split = "[ˆ0-9A-Za-z]"))

moby_dick2 <- unlist(strsplit(moby_dick, split = " ", fixed = TRUE))

Then the second step of selecting the non-blank words would not have been
necessary. But regular expressions will be essential going forward, so it was
worthwhile using regular expressions even if they do not provide the most
efficient method.)

11.3.3 Practice Problem

Use strsplit() and regular expressions to split the following strings into
their respective words (i.e. write a regular expression that will match the -
and . character). Your output should be a vector (not a list).

strings <- c("Once-upon", "a.time", "there", "was-a", "man.named", "Bob")

11.3.4 nchar Again

Now that the vector moby_dick contains each word in the novel as a separate
element, it is relatively easy to do some basic analyses. For example the nchar
function can give us a count of the number of characters in each element of
the vector, i.e., can give us the number of letters in each word in the novel.
11.3 More String Processing Functions 255

moby_dick_nchar <- nchar(moby_dick)

moby_dick_nchar[1:50]

[1] 7 1 8 4 2 7 4 5 3 5 4 3 4 9 6 6 2
[18] 2 5 2 2 5 3 7 10 2 8 2 2 5 1 7 1 5
[35] 4 5 1 6 3 3 3 6 4 2 3 5 2 2 1 3

max(moby_dick_nchar)

[1] 20

ggplot(data = data.frame(nwords = moby_dick_nchar), aes(x = nwords)) +

geom_histogram(binwidth = 1, color = "black", fill = "white") +
ggtitle("Number of letters in words in Moby Dick") +
theme_bw()

Number of letters in words in Moby Dick

50000

40000

30000
count

20000

10000

0 5 10 15 20
nwords

moby_dick_word_table <- table(moby_dick)

moby_dick_word_table <- sort(moby_dick_word_table, decreasing = TRUE)
moby_dick_word_table[1:50]

moby_dick
the of and a to in that it his
256 11 Text Data

14175 6469 6325 4636 4539 4077 3045 2497 2495

i he but s as with is was for
2114 1876 1805 1737 1720 1692 1690 1627 1593
all this at by whale not from him so
1515 1382 1304 1175 1150 1142 1072 1058 1053
on be one you there now had have or
1040 1032 907 884 854 779 767 754 689
were they like me then which what some their
677 649 639 630 628 625 611 608 604
when an are my no
601 590 587 587 581

11.3.5 Practice Problem

The goal of this exercise is to generate a wordcloud image of the 50 most

frqeuent words in Moby Dick that we identifid above. To do this, we need the
wordcloud package (you may also need to install the package tm). Look at
the R documentation for the wordcloud function included in the wordcloud
package and create a wordcloud image of the most popular 500 words in the
novel. Here is what mine looks like
Warning in wordcloud(md.data.frame$moby_dick,
md.data.frame$Freq, max.words = 500, : the could not be
fit on page. It will not be plotted.

that world
green gone close says
young
book queequeg
with
thou
flask here
therefore him may deep
towards part since mate full another sleep within
far certain back leviathan seamen seas
quarter side pequod seemed turned forward having
god

indian

but next
arm who known must
seem tashtego common better
round mast set mr would anything earth mighty called thought suddenly
well
said

down high true

for make day should
little touching jonah means standing ye
chapter bows

heard room
whale

hard after hand

speak second

rope particular
right

black hold
up till taken form themselves them men
ocean

below times
first
two all often nevertheless into upon reason bildad when boats now
his it
feet above man leg made place beneath teeth crew heart
know monster before
fish try her yet my sudden say peleg morning pip whole
sails
was again

to
bone whatever
best five jaw live went strange head voice
mine sail got came however watch
half do never craft no voyage eyes
chase instant large indeed come king nantucket think than vessel take
at

of
am off nigh sure
told very old mark broken good blood such or making stood pull forth
in
least

has three
aloft

thee
air

hundred heaven main what every fine

bottom heads years
perhaps

only white
could

because lost days be end body

over

any

work struck tail did

between behind tell
ll
its fast left last long most your savage
found nothing less keep night not devil
chance death let whose house entire
thy againstturn took
many
way view st new even hole sir
being

these where mind

things four why own turning open an find you more whether once
fact cut
oil soul story sort still
i this
how

point fire does living

short

t vast
harpoon

seems don if looked by rest

near

cabin soon
enough

will rather
looking

light poor life itself

sea wide much among
noble me fellow himself case been people while dick word put
ivory

our fishery
so rolled

whalemen one also

face thing can mouth ere both besides spout
which ship hands ground
lord

great wild sailor each

as

captain ships especially bed

wind

name through sun top shall too away us land cannot else kept
comes oh lower had going english something coffin about sailors lance ever
see

a and
present might home blue small sperm without then some we ten
seen sight grand
broad d ahab stand stranger get
thousand

their
starbuck hesat
like
used
general sometimes stern together same cried
running hear time
are boy give myself few business carpenter feel
almost just were thus whales
s nor
harpooneers
cook
dark
knew

out peculiar from on matter done

goes

other

whaleman
order
boat under have sharks those
stubb yes eye saw run slowly aye taking curious
coming call lay waves door though
line moment straight go dead iron several
began further always
waters along look
bones she board there they
harpooneer length water
deck moby iswhaling
11.3 More String Processing Functions 257

11.3.6 substr and strtrim

The substr function can be used to extract or replace substrings. The first
argument is the string to be manipulated, and the second and third arguments
specify the first and last elements of the string to be extracted or to be replaced.

x <- "Michigan"
substr(x, 3, 4)

[1] "ch"

substr(x, 3, 4) <- "CH"

x

[1] "MiCHigan"

x <- c("Ohio", "Michigan", "Illinois", "Wisconsin")

substr(x, 2,4)

[1] "hio" "ich" "lli" "isc"

substr(x, 2, 4) <- "$#&"

x

[1] "O$#&" "M$#&igan" "I$#&nois" "W$#&onsin"

The strtrim function trims a character string to a specified length.

strtrim("Michigan", 1)

[1] "M"

strtrim("Michigan", 4)

[1] "Mich"

strtrim("Michigan", 100)

[1] "Michigan"

strtrim(c("Ohio", "Michigan", "Illinois", "Wisconsin"), 3)

258 11 Text Data

[1] "Ohi" "Mic" "Ill" "Wis"

strtrim(c("Ohio", "Michigan", "Illinois", "Wisconsin"), c(3, 4, 5, 6))

[1] "Ohi" "Mich" "Illin" "Wiscon"

11.3.7 grep and Related Functions

The grep function searches for a specified pattern and returns either the
locations where this pattern is found or the selected elements. The grepl
function returns a logical vector rather than locations of elements.
Here are some examples. All use fixed = TRUE since at this point we are using
fixed character strings rather than regular expressions.

grep("a", c("the rain", "in Spain stays mainly", "in", "the plain"),
fixed = TRUE)

[1] 1 2 4

grep("a", c("the rain", "in Spain stays mainly", "in", "the plain"),
fixed = TRUE, value = TRUE)

[1] "the rain"

[2] "in Spain stays mainly"
[3] "the plain"

grepl("a", c("the rain", "in Spain stays mainly", "in", "the plain"),
fixed = TRUE)

[1] TRUE TRUE FALSE TRUE

The sub and gsub functions replace a specified character string. The sub
function only replaces the first occurrence, while gsub replaces all occurrences.

gsub("a", "?", c("the rain", "in Spain stays mainly", "in",

"the plain"), fixed = TRUE)

[1] "the r?in"

[2] "in Sp?in st?ys m?inly"
[3] "in"
[4] "the pl?in"
11.4 Exercises 259

sub("a", "?", c("the rain", "in Spain stays mainly", "in",

"the plain"), fixed = TRUE)

[1] "the r?in"

[2] "in Sp?in stays mainly"
[3] "in"
[4] "the pl?in"

gsub("a", "???", c("the rain", "in Spain stays mainly", "in",

"the plain"), fixed = TRUE)

[1] "the r???in"

[2] "in Sp???in st???ys m???inly"
[3] "in"
[4] "the pl???in"

sub("a", "???", c("the rain", "in Spain stays mainly", "in",

"the plain"), fixed = TRUE)

[1] "the r???in"

[2] "in Sp???in stays mainly"
[3] "in"
[4] "the pl???in"

11.4 Exercises

Exercise Text Data Learning objectives: read and write text data; concate-
nate text with the paste function, analyze text with nchar; practice with
functions; manipulate strings with substr and strtrim.
 12
Rcpp

R is an interpreted language. The code interpretor runs each line of a script

individually instead of compiling the entire file at once as in a compiled
language like C++. This means R can be slower than compiled languages
when completing some tasks. In this chapter, we show how to improve the
performance of our R scripts by rewriting functions in C++1 . Although R has
an API for developers to move between R data structures and data structures
in lower-level languages (C++, C, FORTRAN, etc.), an alternative is to use
Rcpp. According to the official Rcpp website2 , “Rcpp provides a powerful API
on top of R, permitting direct interchange of rich R objects between R and
C++.” It is a higher-level approach than using the aforementioned API directly.
C++ functions can address gaps in R’s abilities including
• Vectorizing loops whose subsequent iterations depend on previous iterations.
• Writing recursive functions with lower overhead. (Recursive functions are
functions that call themselves, leading to millions of function calls.)
• Using data structures and algorithms that R does not provide but that are
in the C++ standard template library (STL) such as ordered maps and
double-ended queues.
In this chapter, we will cover an abbreviated/modified version of Hadley
Wickham’s tutorial High performance functions with Rcpp3 .

12.1 Getting Started with Rcpp

12.1.1 Installation

Before we install the Rcpp package, we need a working C++ compiler. To get
the compiler:
1 You do not need prior C++ experience to complete this chapter and corresponding

exercises, but we are hoping you learn a little C++ syntax along the way!
2 http://www.rcpp.org/
3 http://adv-r.had.co.nz/Rcpp.html

261
 262 12 Rcpp

• Windows: install Rtools from here4 .

• Mac: install Xcode from the Mac App Store.
• Linux: in the terminal (for Debian-based systems or similar) enter sudo
apt-get install r-base dev.
Now we can install Rcpp. (Note that the microbenchmark package is needed
for the examples in this chapter, so we’ll install that, too.)

12.1.2 The Simplest C++ Example

It is common for textbooks to introduce a new programming language through

the “Hello World” program.5 If we wanted to write this simple program in
C++, the code would look like this:

#include <iostream>

int main()
{
std::cout << "Hello World!" << std::endl;
}

Think about the C++ code above. How would you accomplish the same task
in R? How are the programs similar in syntax and structure? How are they
different? In the following section, we will rewrite this program in a format that
can interface with R through Rcpp. Note that C++ is a compiled language,
which means we cannot run the program line-by-line like we can with an R
script.

12.2 Using Rcpp

12.2.1 Exporting C++ Functions

Let’s start out with that simple, obligatory “Hello World!” program in C++
that we will export to R. C++ files use the extension *.cpp. Through the
RStudio menu, create a new C++ File (File > New File > C++ File). Note
that the default C++ file contains the line #include <Rcpp.h>. In C++,
these include files are header files that provide us with objects and functions
4 https://cran.r-project.org/bin/windows/Rtools/
5 For a collection of “Hello World” programs in 400+ programming languages, see here6 .
12.2 Using Rcpp 263

that we can use within our C++ file. Although the program building process is
different, there are some parallels between these header files and the packages
we use in our R code. Note that unlike R, every line of code in a C++ file
must end with a semicolon.
To verify that Rcpp is working correctly, let’s run the following code. Save the
document as hello.cpp, and enter the following code:

#include <Rcpp.h>
using namespace Rcpp;

// [[Rcpp::export]]

void hello(){
Rprintf("Hello World!\n");
}

Let’s do a quick line-by line explanation:

1. #include <Rcpp.h>: contains the definitions used by the Rcpp

package.
2. using namespace Rcpp;: allows us to call functions from
the Rcpp header file by writing functionname instead of
Rcpp::functionname.
3. // [[Rcpp::export]]: an Rcpp attribute. Attributes are annota-
tions that are added to C++ source files to indicate that C++
functions should be made available as R functions.
4. void hello(){: the declaration of the hello function, which has a
return type void and no arguments, hence the empty parentheses.
Note that the function content (definition) is all contained within
the squiggly brackets.
5. Rprintf("Hello World!\n");: prints to the R console.

Now, we can go to the RStudio Console and test our code with the following
three lines and the expected output.

library(Rcpp)
sourceCpp("hello.cpp")
hello()

[1] "Hello"
The sourceCpp function parsed our C++ file (“hello.cpp”) and looked for
functions marked with the Rcpp::export attribute. A shared library was then
built and the exported function (hello()) was made available as an R function
264 12 Rcpp

in the current environment. The function simply printed “Hello World!” to our
RStudio Console. Great! You have now written a C++ function using built-in
C++ types and Rcpp wrapper types and then sourced them like an R script.

12.2.2 Inline C++ Code

Maintaining C++ code in it’s own source file provides several benefits including
the ability to use C++ aware text-editing tools and straightforward mapping
of compilation errors to lines in the source file. However, it’s also possible to
do inline declaration and execution of C++ code.
There are several ways to accomplish this, including passing a code string to
sourceCpp() or using the shorter-form cppFunction() or evalCpp() func-
tions. Run the following code:

library(Rcpp)

cppFunction("
int add(int x, int y, int z) {
int sum = x + y + z;
return sum;
}"
)

add(1, 2, 4)

[1] 7

a <- add(3, 2, 1)
a

[1] 6
When you run this code, Rcpp will compile the C++ code and construct an R
function that connects to the compiled C++ function. We’re going to use this
simple interface to learn how to write C++. C++ is a large language, and
there’s no way to cover it all in the limited time we have. Instead, you’ll get
the basics so that you can start writing useful functions to address bottlenecks
in your R code.
 12.4 The Rcpp Interface 265

12.3 The Rcpp Interface

The following subsections will teach you the basics by translating simple R
functions to their C++ equivalents. We’ll start simple with a function that has
no inputs and a scalar output, and then get progressively more complicated:
• No input and scalar output
• Scalar input and scalar output
• Vector input and scalar output
• Vector input and vector output
• Matrix input and vector output
• Matrix input and matrix output

12.4 No input and scalar output

Let’s start with a very simple function. It has no arguments and always returns
the integer 1:

one <- function() 1L

The equivalent C++ function is:

int one() {
return 1;
}

We can compile and use this from R with cppFunction:

cppFunction("
int one() {
return 1;
}"
)

This small function illustrates a number of important differences between R

and C++:
• The syntax to create a function looks like the syntax to call a function; you
don’t use assignment to create functions as you do in R.
 266 12 Rcpp

• You must declare the type of output the function returns. This function
returns an int (a scalar integer). The classes for the most common types of R
vectors are: NumericVector, IntegerVector, CharacterVector, and LogicalVec-
tor.
• Scalars and vectors are different. The scalar equivalents of numeric, integer,
character, and logical vectors are: double, int, String, and bool.
• You must use an explicit return statement to return a value from a function.
• Every statement is terminated by a semicolon.

12.4.1 Scalar input and scalar output

The next example function implements a scalar version of the sign() function
which returns 1 if the input is positive, and -1 if it’s negative:

signR <- function(x) {

if (x > 0) {
1
} else if (x == 0) {
0
} else {
-1
}
}

cppFunction("
int signC(int x) {
if (x > 0) {
return 1;
} else if (x == 0) {
return 0;
} else {
return -1;
}
}"
)

In the C++ version:

• We declare the type of each input in the same way we declare the type of
the output. While this makes the code a little more verbose, it also makes it
very obvious what type of input the function needs.
• The if syntax is identical. While there are some big differences between R
and C++, there are also lots of similarities! C++ also has a while statement
 12.4 No input and scalar output 267

that works the same way as R’s, e.g., as in R you can use break to exit the
loop.

12.4.2 Vector input and scalar output

One big difference between R and C++ is that the cost of loops is much lower
in C++. For example, we could implement the sum() function in R using a
loop. If you’ve been programming in R a while, you’ll probably have a visceral
reaction to this function!

sumR <- function(x) {

total <- 0
for (i in 1:length(x)) {
total <- total + x[i]
}
total
}

In C++, loops have very little overhead, so it’s fine to use them.

cppFunction("
double sumC(NumericVector x) {
int n = x.size();
double total = 0;
for(int i = 0; i < n; i++) {
total += x[i];
}
return total;
}"
)

The C++ version is similar, but

• To find the length of the vector, we use the .size() method, which returns
an integer. C++ methods are called with . (i.e., a period).
• The for statement has a different syntax: for(init; check; increment).
This loop is initialized by creating a new variable called i with value 0.
Before each iteration, we check that i < n, and terminate the loop if it’s not.
After each iteration, we increment the value of i by one, using the special
prefix operator ++ which increases the value of i by 1.
• In C++, vector indices start at 0. I’ll say this again because it’s so important:
IN C++, VECTOR INDICES START AT 0! This is a very common source
of bugs when converting R functions to C++.
 268 12 Rcpp

• Use = for assignment, not <-.

• C++ provides operators that modify in-place: total += x[i] is equivalent
to total = total + x[i]. Similar in-place operators are -=, *=;, and /=.
This is a good example of where C++ is much more efficient than R. As shown
by the following microbenchmark, sumC() is competitive with the built-in (and
highly optimized) sum(), while sumR() is several orders of magnitude slower.
Note, you’ll need to install the microbenchmark package.

library(microbenchmark)

x <- runif(1000)
microbenchmark(sum(x), sumC(x), sumR(x))

Unit: nanoseconds
expr min lq mean median uq max neval
sum(x) 977 1003 1127 1034 1066 8803 100
sumC(x) 2314 2402 12831 2459 2532 1014755 100
sumR(x) 27797 27906 56479 27992 28153 2770329 100
cld
a
a
a
The microbenchmark() function runs each function it’s passed 100 times and
provides summary statistics for the multiple execution times. This is a very
handy tool for testing various function implementations (as illustrated above).

12.4.3 Vector input and vector output

Next we’ll create a function that computes the Euclidean distance between a
value and a vector of values:

pdistR <- function(x, ys) {

sqrt((x - ys)ˆ2)
}

It’s not obvious that we want x to be a scalar from the function definition.
We’d need to make that clear in the documentation. That’s not a problem in
the C++ version because we have to be explicit about types:

cppFunction("
NumericVector pdistC(double x, NumericVector ys) {
 12.4 No input and scalar output 269

int n = ys.size();
NumericVector out(n);

for(int i = 0; i < n; i++) {

out[i] = sqrt(pow(ys[i] - x, 2.0));
}
return out;
}"
)

This function introduces only a few new concepts:

• We create a new numeric vector of length n with a constructor:
NumericVector out(n). This and similar vector and matrix constrictors
initialize the elements with zeros. Another useful way of making a vector is
to copy an existing one: NumericVector zs = clone(ys).
• C++ uses pow(), not ˆ, for exponentiation.
Note that because the R version is fully vectorized, it’s already going to be
fast. On my computer, it takes around 8 ms with a 1 million element y vector.
The C++ function is twice as fast, ∼ 4 ms, but assuming it took you 10
minutes to write the C++ function, you’d need to run it ∼ 150,000 times to
make rewriting worthwhile :-) The reason why the C++ function is faster is
subtle, and relates to memory management. The R version needs to create an
intermediate vector the same length as y(x - ys), and allocating memory is
an expensive operation. The C++ function avoids this overhead because it
uses an intermediate scalar.

12.4.4 Matrix input and vector output

Each vector type has a matrix equivalent: NumericMatrix, IntegerMatrix,

CharacterMatrix, and LogicalMatrix. Using them is straightforward. For
example, we could create a function that reproduces rowSums():

cppFunction("
NumericVector rowSumsC(NumericMatrix x) {
int nrow = x.nrow(), ncol = x.ncol();
NumericVector out(nrow);

for (int i = 0; i < nrow; i++) {

double total = 0;
for (int j = 0; j < ncol; j++) {
total += x(i, j);
270 12 Rcpp

}
out[i] = total;
}
return out;
}"
)

x <- matrix(sample(100), 10)

rowSums(x)

[1] 603 548 514 472 333 564 520 461 566 469

rowSumsC(x)

[1] 603 548 514 472 333 564 520 461 566 469
Let’s look at the main differences:
• In C++, you subset a matrix with (), not [].
• In C++, use .nrow() and .ncol() methods to get the dimensions of a
matrix.

12.4.5 Matrix input and matrix output

See the Rcpp exercise.

12.5 Exercises
Exercise Rcpp Learning objectives: practice using Rcpp to run a C++ func-
tion through R; use microbenchmark() to compare function performance.
13
Databases and R

As we previously saw in Chapter 6, dplyr is a fantastic tool for manipulating

data inside the R environment. Up to this point, we have worked with data
sets that easily fit within your computer’s memory. Is R capable of working
with data sets that are too large to store on your computer? Fortunately R is
more than capable of dealing with large data sets (or not so large data sets
that are stored remotely in a database). Recently, developers at RStudio have
been working on building tools for data scientists to use inside of R to work
with databases. Currently there are numerous different ways to work with data
from databases in R. In this chapter we will explore some of the many different
packages used to query external databases from within R, and we will focus
on using dplyr to perform our queries solely in R.

13.1 SQL and Database Structure

In today’s world, there is an extremely large amount of electronic data. The
actual amount of existing data is not quite known, as it is pretty much
impossible to determine an exact amount. One estimate by IBM stated that
2.5 exabytes (2.5 billion gigabytes) of data were generated every day in 2012.
Clearly, we need to have efficient ways to store such incredibly large amounts
of data inside computers. In many cases, these data are stored in relational
databases.
In simple terms, a database is a collection of similar files. In a relational
database, there are numerous data-containing tables that are related to each
other by some common field, known as a key. Each table consists of numerous
rows and columns, very similar to the way a tibble or any other type of data
frame is stored inside of R. One can imagine a database consisting of student
records with two tables, one containing the address information of students
(called address), and one containing email information of the students (called
email). These two tables could perhaps be linked (or joined) together by a
common field such as a student identification number. Thus one could join

271
272 13 Databases and R

together the two tables to obtain email information and address information
for any given student at the same time.
Data analysts and data scientists require a method to do tasks like that
described above in databases. Fortunately, almost every relational database is
manipulated and stored using SQL, which makes such queries possible.

13.1.1 SQL

SQL1 , or Structured Query Language, is an incredibly useful tool for querying

and managing relational databases. SQL is common to all major database man-
agement systems, such as Oracle, MariaDB, PostgreSQL, MySQL, SQLite, and
SQL Server. Each system has slight differences in syntax for certain commands,
but the overall basic structure is the same across all database management
systems. In order to work with databases, at least a basic understanding of
the SQL language is desirable. Here we will give a brief overview of the main
structure of a SQL query. If you have minimal SQL background and desire to
learn more about databases, we encourage you to use numerous free online
resources to help gain experience in using SQL. Some of my favorites include
tutorials from W3Schools2 , tutorialspoint3 , and SQL Tutorial4 . There are also
numerous free online courses available (i.e. this Stanford5 course) that you can
enroll in and gain experience in querying on sample databases.

13.1.1.1 A Basic Query

Let’s first explore a little bit of SQL syntax. A basic SQL query consists of
three different parts:

1. SELECT: used to select data from a database

2. FROM: tells SQL what table to look at to get the data
3. WHERE: used to filter out different records

Using these three statements you are able to perform queries on the database
to extract data from different tables. For example, consider the database
mentioned previously consisting of student records. To obtain all the email
addresses from the email table, our query would look like this
1 The pronunciation of SQL is highly debated. I prefer to pronounce it

like the word “sequel”, but many others prefer to say it as “ess-que-ell”.
See https://softwareengineering.stackexchange.com/questions/8588/whats-the-hist
ory-of-the-non-official-pronunciation-of-sql for a good debate
2 https://www.w3schools.com/sql/
3 https://www.tutorialspoint.com/sql/index.htm
4 http://www.sqltutorial.org/
5 https://lagunita.stanford.edu/courses/Engineering/db/2014_1/about
13.2 SQL and Database Structure 273

SELECT email_address
FROM email
If we only desired the email addresses of the student with id = 001, then we
would add a condition using the WHERE clause:
SELECT email_address
FROM email
WHERE student_id = '001'
Queries of this basic form are used by database programmers numerous times
a day in order to obtain needed information from large databases. There are
many more features that can be added to this basic query form, including:

1. AND: added to the WHERE clause, allows for multiple conditions at

once
2. JOIN: connect two tables based on a common feature
3. GROUP BY: group data together according to a certain field to obtain
statistics, counts, etc.

Does this remind you of anything we previously studied? Maybe something in

Chapter 6? If so, you are correct! The manipulation of data sets in R using
dplyr provides many of the same sort of manipulation tools used by SQL, and
we will soon see how we can use dplyr directly to query external databases.
SQL also has the ability to update the database, insert new records, and delete
records from the database. This is done using the UPDATE, INSERT, and DELETE
statements. Here are some simple examples for each of these statements:
UPDATE email
SET email_address = 'doserjef@msu.edu'
WHERE student_id = '001'

DELETE FROM email

WHERE email_address = 'doserjef@msu.edu'

INSERT INTO email

VALUES ('001', 'doserjef@msu.edu')
Again, this is nowhere near a full introduction to the SQL language, but it will
provide you with enough understanding to work with databases within the R
framework. If you are interested in databases, we encourage you to seek out
some of the additional resources mentioned at the beginning of this section.
274 13 Databases and R

13.2 Difficulties of Working with Large Datasets

The first step in using R with databases is to connect R to the database. This
leads to many difficulties. As described by Edgar Ruiz, a solutions engineer at
RStudio, when connecting to a database in R there is only a “small conduit”
between R and the database that often prevents us from being able to work
with the data as fast as we can work with local data frames. In order to work
around this there have historically been two different options. The first option
is to take data from the database in small chunks, save it in R memory, and
then analyze it later. This is not desirable because you aren’t dealing with the
data directly, so if the data in the database changes this will not be reflected
in the data previously loaded into your R memory. A different option is to
load all of the data into R. This allows you to see all of the data, but it takes
forever to download and it essentially takes over the R session (and it has the
same problems as option one regarding changes in the database).
The main problem with these methods is that they are trying to perform the
computations locally in R as opposed to performing computations directly on
the data in the database using the SQL Engine. Ideally, we want to be able to
analyze the data in place. In other words, we want R to somehow send a SQL
query over to the database, perform the query using the powerful SQL engine,
and then have the database send back an R data frame that we can manipulate
in R. This would allow us to avoid performing computations directly in R,
which would improve program speed and reduce the amount of storage needed.
In addition, we do not want to go back and forth between using R and SQL,
as this can cause a lot of unnecessary confusion. As R programmers, we of
course want to write solely R code. Fortunately, researchers at RStudio have
increased the capabilities of the dplyr package to allow just that.

13.3 Using dplyr to Query the Database

The dplyr package has recently been updated to allow for better communi-
cations with external databases. Specifically, you, as the R programmer, can
write code in dplyr acting on a database in the same manner in which you
use dplyr to work with a data frame in the R environment. dplyr will then
behind the scenes convert this R code into a SQL query, will send this query
to the database, use the SQL engine to run the query on the data, and will
return these data as a data frame in R.
This is the most efficient method for querying databases in R for three reasons:
13.3 Using dplyr to Query the Database 275

1. We are pushing the computation to the database, allowing us to

avoid bringing large amounts of data in the database into the R
environment unless we specifically ask R to do it.
2. We can use piped code, which greatly enhances the readability of
the code.
3. All the code is in R. This reduces the cognitive costs that are often
associated when using two or more programming languages at the
same time.

Next we will go through an example using SQLite to detail exactly how queries
on a sample database are performed.

13.3.1 Example with RSQLite

SQLite is arguably the most widely used relational database management

system throughout the world. We will first provide a demonstration using a
SQLite database as it is relatively easy to install on your system compared to
other database management systems. SQLite is different from most database
management systems because you don’t have to set up a separate database
server. This makes SQLite great for a demo, and surprisingly it can also
be very powerful when working with many gigabytes of data. We will use a
sample database provided by SQLite called chinook that contains data on
store employees, customers, and the music they purchased.
First, go to http://www.sqlitetutorial.net/sqlite-sample-database/
and download the chinook database into your R working directory so you can
follow along with this example yourself.
Upon successful download, the first step we need to do is install/load
the necessary packages for connecting to the external database. To help
dplyr communicate with the SQL engine, install the dbplyr package
using install.packages("dbplyr"). Next we need the RSQLite pack-
age for interfacing with the SQLite database. Install this by running
install.packages("RSQLite") in the R console. Then load the packages,
along with dplyr:

library(dplyr)
library(dbplyr)
library(RSQLite)

We next use the DBI package to connect directly to the database. DBI is
a backend package that provides a common interface for R to work with
many different database management systems using the same code. This
package does much of the communication from R to the database that occurs
276 13 Databases and R

behind the scenes, and is an essential part of using dplyr to work with
databases. If you have not yet installed the DBI package, you can do so using
install.packages("DBI").

library(DBI)
chinook <- dbConnect(SQLite(), "chinook.db")

This command creates a reference to the database, and tells R to connect

to this database in a specific location (your location could change depending
on where you save the chinook.db file). A similar approach could be used
to connect to databases of other database management systems like Oracle,
MySQL, PostgreSQL, and others. Most databases do not live in a file, but
instead live on another server. This causes the above connection to be much
more complex, but for now we will focus on this simple case.
Now lets look closer at the chinook database.

src_dbi(chinook)

src: sqlite 3.30.1 [/home/jeffdoser/Dropbox/teaching/for875-20/textbook/chinook.db]

tbls: albums, artists, customers, employees, genres,
invoice_items, invoices, media_types,
playlist_track, playlists, sqlite_sequence,
sqlite_stat1, tracks
The above function displays the location of the database, as well as the tables
contained in the database. You can see in the chinook database there are
numerous tables regarding customers, employees, and the music that customers
purchased. In total, there are 13 tables contained in this database.
Now that we know the structure of the database, we can perform some simple
queries on the data using dplyr syntax. For now, let’s focus on the employees
table.

employees <- tbl(chinook, "employees")

employees

# Source: table<employees> [?? x 15]

# Database: sqlite 3.30.1
# [/home/jeffdoser/Dropbox/teaching/for875-20/textbook/chinook.db]
EmployeeId LastName FirstName Title ReportsTo
<int> <chr> <chr> <chr> <int>
1 1 Adams Andrew Gene~ NA
2 2 Edwards Nancy Sale~ 1
3 3 Peacock Jane Sale~ 2
13.3 Using dplyr to Query the Database 277

4 4 Park Margaret Sale~ 2

5 5 Johnson Steve Sale~ 2
6 6 Mitchell Michael IT M~ 1
7 7 King Robert IT S~ 6
8 8 Callahan Laura IT S~ 6
# ... with 10 more variables: BirthDate <chr>,
# HireDate <chr>, Address <chr>, City <chr>,
# State <chr>, Country <chr>, PostalCode <chr>,
# Phone <chr>, Fax <chr>, Email <chr>
Notice how the employees table looks mostly like a regular tibble, but has a
couple added lines detailing its location as a remote table in a SQLite database.
Now lets use familiar dplyr commands to perform queries on the database6 .

employees %>%
select(LastName, FirstName, Phone, Email) %>%
arrange(LastName)

# Source: lazy query [?? x 4]

# Database: sqlite 3.30.1
# [/home/jeffdoser/Dropbox/teaching/for875-20/textbook/chinook.db]
# Ordered by: LastName
LastName FirstName Phone Email
<chr> <chr> <chr> <chr>
1 Adams Andrew +1 (780) 428-~ andrew@chinookcorp~
2 Callahan Laura +1 (403) 467-~ laura@chinookcorp.~
3 Edwards Nancy +1 (403) 262-~ nancy@chinookcorp.~
4 Johnson Steve 1 (780) 836-9~ steve@chinookcorp.~
5 King Robert +1 (403) 456-~ robert@chinookcorp~
6 Mitchell Michael +1 (403) 246-~ michael@chinookcor~
7 Park Margaret +1 (403) 263-~ margaret@chinookco~
8 Peacock Jane +1 (403) 262-~ jane@chinookcorp.c~

employees %>%
filter(Title == "Sales Support Agent") %>%
select(LastName, FirstName, Address) %>%
arrange(LastName)

# Source: lazy query [?? x 3]

# Database: sqlite 3.30.1
# [/home/jeffdoser/Dropbox/teaching/for875-20/textbook/chinook.db]
# Ordered by: LastName
6 You may see a warning here. You can safely ignore it.
278 13 Databases and R

LastName FirstName Address

<chr> <chr> <chr>
1 Johnson Steve 7727B 41 Ave
2 Park Margaret 683 10 Street SW
3 Peacock Jane 1111 6 Ave SW

employees %>%
group_by(ReportsTo) %>%
summarize(numberAtLocation = n())

# Source: lazy query [?? x 2]

# Database: sqlite 3.30.1
# [/home/jeffdoser/Dropbox/teaching/for875-20/textbook/chinook.db]
ReportsTo numberAtLocation
<int> <int>
1 NA 1
2 1 2
3 2 3
4 6 2
You can see by using dplyr you are able to use the same syntax to query
the database as you used to analyze data frames in Chapter 6. The most
important difference to again note is that when working with remote databases
the R code is translated into SQL and executed in the database using the SQL
engine, not in R. When doing this, dplyr is as “lazy” as possible as it never
pulls data in R unless explicity asked. It collects everything you ask it to do
and then sends it to the database all in one step. This is often a very useful
feature when desiring to look at a certain group of records in a database. To
understand this further, take a look at the following code, which at first you
might think will output a tibble or data frame:

salesSupportAgents <- employees %>%

filter(Title == "Sales Support Agent") %>%
select(LastName, FirstName, Address) %>%
arrange(LastName)
salesSupportAgents

# Source: lazy query [?? x 3]

# Database: sqlite 3.30.1
# [/home/jeffdoser/Dropbox/teaching/for875-20/textbook/chinook.db]
# Ordered by: LastName
LastName FirstName Address
<chr> <chr> <chr>
1 Johnson Steve 7727B 41 Ave
2 Park Margaret 683 10 Street SW
13.3 Using dplyr to Query the Database 279

3 Peacock Jane 1111 6 Ave SW

Notice the first two lines in the output. The source is described as a lazy
query and the salesSupportAgents is still a database connection and not a
local data frame. Because you did not specifically tell R that you wanted to
bring the data directly into R, it did not do so. In order to bring the data
directly into R in a local tibble you need to use the collect() function. Note
that if you are asking the database to send back a lot of data, this could take
a while.

salesSupportAgents <- employees %>%

filter(Title == "Sales Support Agent") %>%
select(LastName, FirstName, Address) %>%
arrange(LastName) %>%
collect()
salesSupportAgents

# A tibble: 3 x 3
LastName FirstName Address
<chr> <chr> <chr>
1 Johnson Steve 7727B 41 Ave
2 Park Margaret 683 10 Street SW
3 Peacock Jane 1111 6 Ave SW
As we’ve seen, for simple tasks, and even many complex tasks, dplyr syntax
can be used to query external databases.

13.3.2 dbplot

If we can use dplyr to analyze the data in a database, you may be wondering
whether or not we can use ggplot2 to graph the data in the database. Of course
we can! In fact, the package dbplot is designed to process the calculations of
a plot inside a database and output a ggplot2 object. If not already installed
on your system, make sure to install the dbplot package before continuing.
We can use the same chinook database from SQLite we were using above.
Suppose we desire to see how many types of each employee there are in the
database. We can produce a barplot to show this.

library(dbplot)
employees %>%
dbplot_bar(Title)
280 13 Databases and R

2
n ()

General Manager IT Manager IT Staff Sales Manager Sales Support Agent

Title

We first load the dbplot package. Next we produce the bar plot. Notice that we
can continue to use the convenient %>% character while producing graphs using
dbplot, making the code easy to read. Since dbplot outputs a ggplot2 object,
we can further customize the graph using familiar functions from ggplot2 (so
long as the package is loaded).

library(ggplot2)
employees %>%
dbplot_bar(Title) +
labs(title = "Employee Types") +
ylab("Count") +
theme_classic()
13.3 Using dplyr to Query the Database 281

Employee Types
3

2
Count

General Manager IT Manager IT Staff Sales Manager Sales Support Agent

Title

13.3.3 Using Google BigQuery

For a second example, we will utilize a database from Google BigQuery, Google’s
fully managed low cost analytics data warehouse. The package bigrquery
provides an R interface to Google BigQuery. If not already installed, install
the package with our usual method.
We will use data from a great set of sample tables that BigQuery provides to
its users. First we will look at the shakespeare table that contains a word
index of the different works of Shakespeare. Thus, given the data in the table,
let’s use bigrquery, DBI, and dplyr to determine the ten words that appear
most often in Shakespeare’s works.

library(bigrquery)
library(DBI)
library(dplyr)
billing <- "for875-databases"
con <- dbConnect(
bigquery(),
project = "publicdata",
dataset = "samples",
billing = billing
)
282 13 Databases and R

shakespeare <- con %>%

tbl("shakespeare")
shakespeare %>%
group_by(word) %>%
summarise(n = sum(word_count, na.rm = TRUE)) %>%
arrange(desc(n)) %>%
head(10)

# Source: lazy query [?? x 2]

# Database: BigQueryConnection
# Ordered by: desc(n)
word n
<chr> <int>
1 the 25568
2 I 21028
3 and 19649
4 to 17361
5 of 16438
6 a 13409
7 you 12527
8 my 11291
9 in 10589
10 is 8735
First we load the neceessary packages. We then provide the name of our
project to the billing variable. In this case, I created a project called
for875-databases for this course. The con variable establishes the connection
to the database and the data that we want to look at. Notice the difference
in connecting to the Google BigQuery database, which is stored remotely on
Google’s servers, as compared to the previous example when connecting to a
SQLite database stored on your computer’s hard drive. This example is much
more realistic of the connections you would make when accessing external
databases. The bigquery() statement tells DBI we will be working with a
database from Google BigQuery. The project and dataset options tell DBI
what database to look for inside of the Google BigQuery framework. We then
supply the name of our project to the billing option to tell BigQuery what
account to “bill” (this only costs money if you surpass the monthly limit given
by Google).
If you run this code by yourself, you’ll notice that R will ask you to sign in to a
google account (your MSU account), and then after you allow the bigrquery
package to connect to your account, you will be able to run all the code without
a problem.
From here we can establish a connection to a specific table by using the tbl()
13.3 Using dplyr to Query the Database 283

function as shown above. We can then use dplyr like we did previously to
group the data by words, determine the count of each word, and then order it
in decreasing order to obtain the top ten words in Shakespeare’s works. Notice
that when you run this code you may obtain a message in red describing the
total amount of bytes billed. This is how Google manages how much its users
are querying and using its data. We’re just tinkering around here, so we won’t
surpass the monthly limit on free queries.
This code again displays the benefits of the “lazy evaluation” that R employs.
The con variable and shakespeare variable do not store the database or the
specific table in R itself, instead they serve as references to the database
where the specific data they are referencing is contained. Only when the dplyr
query is written and called using the collect() function is any data from the
database actually brought into R. R waits till the last possible second (i.e. is
lazy) to perform any computations.
A second table in the publicdata project on Google BigQuery contains weather
data from NOAA ranging from the years 1929 to 2010. Let’s first make a
connection to this table and explore its structure.

weather <- con %>%

tbl("gsod")
dbListFields(con, "gsod")

[1] "station_number"
[2] "wban_number"
[3] "year"
[4] "month"
[5] "day"
[6] "mean_temp"
[7] "num_mean_temp_samples"
[8] "mean_dew_point"
[9] "num_mean_dew_point_samples"
[10] "mean_sealevel_pressure"
[11] "num_mean_sealevel_pressure_samples"
[12] "mean_station_pressure"
[13] "num_mean_station_pressure_samples"
[14] "mean_visibility"
[15] "num_mean_visibility_samples"
[16] "mean_wind_speed"
[17] "num_mean_wind_speed_samples"
[18] "max_sustained_wind_speed"
[19] "max_gust_wind_speed"
[20] "max_temperature"
[21] "max_temperature_explicit"
284 13 Databases and R

[22] "min_temperature"
[23] "min_temperature_explicit"
[24] "total_precipitation"
[25] "snow_depth"
[26] "fog"
[27] "rain"
[28] "snow"
[29] "hail"
[30] "thunder"
[31] "tornado"

weather %>%
select(thunder, mean_wind_speed) %>%
head(10)

# Source: lazy query [?? x 2]

# Database: BigQueryConnection
thunder mean_wind_speed
<lgl> <dbl>
1 FALSE 7
2 FALSE 14.5
3 FALSE 28.7
4 FALSE 15.5
5 FALSE 0
6 FALSE 15
7 FALSE 15
8 FALSE 21
9 FALSE 8.90
10 TRUE 13
First we establish a connection to the weather table. We then use the
dbListFields() function from the DBI package to display the fields in the
table. We then view the first 10 records in the table. We see that there are 31
different variables on information for the weather on a given day from 1929
to 2010. Let’s say we are interested in determining the total number of days
where the mean_temp > 60 and mean_wind_speed > 10.

weather %>%
filter(mean_temp > 60 & mean_wind_speed > 10) %>%
summarize(count = n())

# Source: lazy query [?? x 1]

# Database: BigQueryConnection
count
<int>
13.3 Using dplyr to Query the Database 285

1 6581146
Further, suppose we are interested in determining how the average wind speed
has changed from 1929 to 2010. We can use dbplot to plot the data.

weather %>%
dbplot_line(year, average_wind_speed = mean(mean_wind_speed, na.rm = TRUE))

12
average_wind_speed

10

1940 1960 1980 2000

year

Upon first glance at this plot, a naive student might think “Wow! Wind speeds
have decreased dramatically since the 1930s’ ’, and just accept this as true
since that is what the graph shows. But since we are all data analysts at heart
(otherwise you wouldn’t be taking this course!), we want to explore this further.
We might ask why the wind was so high in the late 1920s and early 1930s?
An average wind speed of above 12 miles per hour seems pretty high. Let’s
explore the entries for all years in the 1930s and 1920s by looking at how many
records there are with a mean wind speed for each year before 1940.

weather %>%
filter(year < 1940 & !is.na(mean_wind_speed)) %>%
group_by(year) %>%
summarize(count = n())

# Source: lazy query [?? x 2]

# Database: BigQueryConnection
year count
286 13 Databases and R

<int> <int>
1 1931 9726
2 1934 20334
3 1935 26829
4 1939 65623
5 1932 10751
6 1936 50514
7 1937 83310
8 1929 2037
9 1930 7101
10 1933 17708
# ... with more rows
Interesting! We see that there are only 2037 records from 1929, while there are
65623 records from 1939. This suggests we could be experiencing a phenomenon
known as sampling bias in which the 2037 records from 1929 are not a valid
random representation of all the wind speeds occurring in 1929. Or potentially
the wind was just much higher during that time due to other factors we aren’t
exploring here. Determining the true cause of this pattern requires further
investigation, which we leave to you if you so desire.

13.4 Changing Records in the Database

dplyr is fantastic for retrieving data from databases and analyzing it using
familiar functions. As detailed in this video7 , using dplyr is considered a
“best practice” when using databases in R because you only have to use
one language, you don’t need to know any SQL, and you can use the lazy
evaluation of R, among many other reasons. As a data analyst, a majority of
work can be accomplished by using dplyr and other forms of queries against
a database. But what if you need to do more than query the data in the
database and you have to change some of the records? dplyr is not capable of
inserting or deleting records in a database, and although updating recordings
is possible with dplyr, Hadley Wickham strongly recommends against it, as
dplyr is designed specifically for retrieving data. Thus, we need to use other
options for updating/inserting/deleting records in a database. Unlike using
dplyr this requires a more in depth knowledge of SQL. We will not discuss
updating/inserting/deleting records in a database here, but you will briefly
learn about methods to do this in this chapter’s exercise.
7 https://www.rstudio.com/resources/webinars/best-practices-for-working-

with-databases/
13.7 R Studio Connections Pane 287

13.5 R Studio Connections Pane

If by this point in the book you haven’t already been convinced that RStudio
is a fantastic IDE for running R then perhaps this will convince you. R Studio
has a Connections Pane by default in the top right corner of your screen that
makes it easy to connect to a variety of data sources, and better yet, explore
the objects and data inside the connection. You can use it with a variety of
different packages for working with databases in R, and it will also work with a
variety of databases and other datasources (i.e. Spark). This Connections Pane
is a great way of exploring the data in a data source once you are connected.
If interested, the page here8 which provides an overview of the connections
pane and some other useful links of potential interest.

13.6 Further Resources

There are numerous other methods besides dplyr to work with databases in
R that involve using more standard SQL queries. If you come from a SQL
background and want to use these other methods, then explore the RStudio
Databases using R page9 . This website will be an essential resource for you if
you get more involved with using databases within the R environment.

13.7 Exercises
Exercise Databases Learning objectives: connect to an external database;
perform simple queries using dplyr; use data from a database in a Shiny app;
learn how to perform changes (update/delete/insert) on a database

8 https://db.rstudio.com/rstudio/connections/
9 http://db.rstudio.com/
14
Digital Signal Processing

In today’s technology age, signals are pervasive throughout most of the world.
Humans and animals use audio signals to convey important information to
conspecifics. Airplanes use signals in the air to obtain important information
to ensure your safety on a flight. Cell phones are pretty much just one small
digital signal processing device. They process our speech when we talk into the
phone by removing background noise and echos that would distort the clearness
of our sound. They obtain wifi signals to allow for us to search the web. They
send text messages using signals. They use digital image processing to take
really good pictures. They take a video of your dog when he’s doing something
completely hilarious that you want to show your friend. The applications
of digital signal processing (DSP) span numerous different fields, such as
ecology, physics, communication, medicine, computer science, mathematics,
and electrical engineering. Thus, having a basic understanding of DSP and it’s
concepts can be very helpful, regardless of the field you decide to pursue.
In this chapter we will give a brief introduction to some central concepts in DSP
and how we can explore these concepts using R. We will look at applications
of DSP using some R packages designed for bioacoustic analysis. When doing
digital signal processing, R is not typically the first computer language that
comes to mind. Matlab, Octave, and C/C++ are often considered to be some
of the best languages for DSP. But R provides numerous different packages
(especially for working with audio data) that make it competitive with these
languages for digital signal processing. We will use the signal package to do
some simple digital signal processing, and then will use the tuneR, warbleR,
and seewave packages to work with audio data to show you more specific
examples of how R can be used in different fields of digital signal processing.

14.1 Introduction to Digital Signal Processing

First let’s go over some basics of DSP before we delve into working with signals
in R. Digital Signal Processing is the use of digital tools to perform different
signal processing operations, such as the analysis, synthesis, and modification

289
290 14 Digital Signal Processing

of signals. But what exactly constitutes a signal? A signal is anything that

carries information. In other words, it is a quantity that varies with time,
temperature, pressure, or any other independent variable. As described above,
signals pervade our everyday life, and thus the study of their behavior, how
they work, how we can manipulate them, and how we can extract information
from them is extremely important.
There are two main types of signals: continuous time signals and discrete time
signals. Continuous Time Signals, as you might suspect, are continuous. They
are defined at all instances of time over a given period. These types of signals
are commonly referred to as analog signals. An example of an analog signal
would be a simple sine curve plotted in 14.1. Notice that the sine curve is
defined for every value on the interval from 0 to 2π. This is an analog signal.

1.0

0.5
sin(x)

0.0

-0.5

-1.0

0 2 4 6
x

FIGURE 14.1: Example of an analog signal: a simple sine curve

Discrete Time Signals are not continuous. They are defined only at discrete
times, and thus they are represented as a sequence of numbers (so by your
experiences in R working with sequences of numbers you should already have
an idea of how discrete signals are stored, how they can be manipulated,
etc. . . ). Figure 14.2 is an example of the exact same sine curve as in Figure
14.1 Introduction to Digital Signal Processing 291

14.1, but in a discrete form. Notice how there are only values at specific points,
and anywhere besides those points the signal is not defined. In a real-world
application, these points would correspond to measurements of the signal.

1.0

0.5
sin(x)

0.0

-0.5

-1.0

0 2 4 6
x

FIGURE 14.2: Example of a discrete signal: a sine curve with some missing
values

Notice the differences between the two graphs. The second graph is not defined
at every point and is only defined at the points where a data point exists.
This type of sampling will become important later in the chapter and in the
exercise when we begin to work with sound data (which by the way takes the
form of these simple sine curves we are working with here). There are many
other shapes and forms of signals, but since we will later be focusing on audio
data, and the sine curve is fundamental to any sort of sound, we for now will
focus on working with a sine curve.
Signals have a wide variety of properties and characteristics that give each
signal distinct behavior. The sine curve has two important characteristics.
First, a sine curve is said to be odd since it satisfies the condition sin(−t) =
-sin(t). This is easy to recognize by looking at the graph of a sine curve as you
see that it is symmetrical over the diagonal at the origin. In addition, a sine
292 14 Digital Signal Processing

curve is a periodic signal. A periodic signal is a signal that repeats itself after
a certain interval of time (called the Period). This is true of all real sine curves.
Understanding the simple properties of signals like we have done with the sine
curve is a useful tool in digital signal processing as it allows you to recognize
simple patterns that may occur with the signal in which you are interested.
Now that we have a general understanding of what a signal is and how we
can use it’s properties to learn about it’s behavior, let’s now focus on some
of the most important concepts in digital signal processing, and how we can
implement them in R. Here are three of the most central topics to DSP:

1. Fourier Transforms: signal processing involves looking at the fre-

quency representation of signals for both insight and computations.
2. Noise: defined as any signal that is not desired. A huge part of
signal processing is understanding how noise is affecting your data
and how you can efficiently remove or manage it.
3. Filters: allow you to remove out specific portions of a signal at once,
or allow you to remove noise from certain signals.

We will first briefly discuss working with noise and filters in R, and then we
will work with Fourier Transforms in the context of audio data.

14.1.1 Sinusoidal Signals

As mentioned previously, sinusoidal signals are extremely important for signal

processing. We saw an example of a continuous sinusoid and a discrete sinusoid.
Here we will quickly discuss the mathematics behind coding a sinuosoid signal.
A sine wave typically takes the following form:

f (θ) = (α)sin(b(θ − c)) + d

The translations that you see above from the standard sine wave (sin(θ)) have
particular terms. Here we give a quick summary of the terms in the above
equation and their definitions:

1. Period: the amount of time required to complete one full cycle of

the sine wave. A standard sine wave has period 2π. The sine wave
described above has period 2π|b|
2. Frequency: the number of times a sine wave repeats itself within
a single unit of time (which depends upon how the variable θ is
defined). This is the reciprocal of the period so in our case this is
|b|
2π .
3. Amplitude: the “height” of the sine wave, or more properly, the
distance from the x-axis to the most positive point of the sine wave.
14.2 Introduction to Digital Signal Processing 293

The standard sine wave has an amplitude of 1. The amplitude is

controlled by the α parameter, and thus our sine wave has amplitude
α.
4. Phase Shift: determines whether or not the sine wave passes
through the point (0, 0). Our sine wave mentioned above is shifted
to the right by the amount c and shifted up by d.

Next we draw a few plots to show graphically how all these parameters influence
the signal. We will graph these as digital signals.

library(dplyr)
x <- seq(0, 2 * pi, .1)
standard <- sin(x)
altered <- 2*sin(3 * (x - 1)) + 4
graphData <- tibble(x, standard, altered)
ggplot(data = graphData, mapping = aes(x = x)) +
geom_point(mapping = aes(y = standard, color = "sin(x)")) +
geom_point(mapping = aes(y = altered, color = "2*sin(3 * (x - 1)) + 4")) +
scale_color_manual("", breaks = c("sin(x)", "2*sin(3 * (x - 1)) + 4"),
values = c("blue", "red")) +
labs(x = "x", y ="")

sin(x)

2 2*sin(3 * (x - 1)) + 4

0 2 4 6
x

Now that we understand the fundamentals of sinusoids, let’s see how we can
work with noise and filters on sinusoids using R.
294 14 Digital Signal Processing

14.2 Noise and Filters

Using filters to remove noise and obtain only the desired portions of signals is
a huge part of signal processing. Here we will give a couple examples of the
many different functions available in the signal package for DSP. The signal
package is a set of signal processing functions that was originally written for
more “standard” signal processing languages, Matlab and Octave, but was
later translated into R. It provides an extensive set of functions including
filtering functions, resampling routines, and interpolation functions. To follow
along yourself, install the package with install.packages("signal").
For a first example, consider the sinusoid generated below.

library(signal)
t <- seq(0, 1, len = 100)
sig <- sin(2 * pi * t)
ggplot(mapping = aes(t, sig)) +
geom_line()

1.0

0.5

0.0
sig

-0.5

-1.0

0.00 0.25 0.50 0.75 1.00

t

This is a simple sinusoid that has an amplitude of 1 and a frequency of 1. It

is not likely that when obtaining a real-world signal you would get a perfect
sinusoid without any noise. Let’s take this graph and add some noise to it.
14.2 Noise and Filters 295

noisySig <- sin(2 * pi * t) + 0.25 * rnorm(length(t))

ggplot() +
geom_line(aes(t, noisySig), color = "red")

1.5

1.0

0.5
noisySig

0.0

-0.5

-1.0

-1.5
0.00 0.25 0.50 0.75 1.00
t

We add noise to the function by adding a random value

(0.25*rnorm(length(t)) to each value in the signal. This is more typ-
ical of a real-world signal, as you can somewhat see the general shape (the
signal you are interested in), but it is somewhat distorted as a result of noise.
The signal package provides a long list of functions for filtering out noise.
Here we will utilize the butter function to remove the noise from this signal
in an attempt to obtain our original signal. This function generates a variable
containing the Butterworth filter polynomial coefficients. The Butterworth
filter is a good all-around filter that is often used in radio frequency filter
applications. We then use the filter function from the signal package to
apply the Butterworth filter to the data. Finally we produce a plot using
ggplot2 that shows the original signal without noise, the noisy signal, and
the recovered signal.

library(dplyr)
butterFilter <- butter(3, 0.1)
recoveredSig <- signal::filter(butterFilter, noisySig)
allSignals <- data.frame(t, sig, noisySig, recoveredSig)
ggplot(allSignals, aes(t)) +
296 14 Digital Signal Processing

geom_line(aes(y = sig, color = "Original")) +

geom_line(aes(y = noisySig, color = "Noisy")) +
geom_line(aes(y = recoveredSig, color = "Recovered")) +
labs(x = "Time", y = "Signal")

1.5

1.0

0.5

colour
Noisy
Signal

0.0
Original
Recovered

-0.5

-1.0

-1.5
0.00 0.25 0.50 0.75 1.00
Time

You can see that the recovered signal is not perfect, as there is still some noise
in the signal, and the timing of the peaks in the signal is not exactly matched
up with the original. But it is clear that we have nonetheless removed a large
portion of the noise from the noisy signal. Mess around with the argument
values in the butter() function in the above code. See what changing the
parameters does to the recovered graph. You can also explore the wide variety
of filter functions available in the signal package by exploring the package
documentation here1 .
For a second example, suppose we are interested in not only the signal, but
also the noise. We will extract the low and high frequencies from a sample
signal. Let’s start with a noisy signal

t <- 1:500
cleanSignal <- 50 * sin(t * 4 * pi/length(t))
noise <- 50 * 1/12 * sin(t * 4 * pi/length(t) * 12)
originalSignal <- cleanSignal + noise

1 https://cran.r-project.org/web/packages/signal/signal.pdf
14.2 Noise and Filters 297

ggplot() +
geom_line(aes(t, originalSignal))

30
originalSignal

-30

0 100 200 300 400 500

t

Again, you can easily recognize the pattern is sinusoidal, but there is noise
in the signal that you want to extract. Perhaps you are interested in where
the noise is coming from and want to analyze the structure of the noise to
figure this out. Thus, unlike the previous example, we want to extract the
noise from the signal and not simply eliminate it. To do this we can again use
the Butterworth filter.

lowButter <- butter(2, 1/50, type = "low")

low <- signal::filter(lowButter, originalSignal)

highButter <- butter(2, 1/25, type = "high")

high <- signal::filter(highButter, originalSignal)

signals <- data.frame(originalSignal, low, high)

ggplot(signals, aes(t)) +
geom_line(aes(y = originalSignal, color = "Original")) +
geom_line(aes(y = low, color = "Signal")) +
geom_line(aes(y = high, color = "Noise")) +
labs(x = "Time", y = "Signal")
298 14 Digital Signal Processing

30

colour
Noise
Signal

0
Original
Signal

-30

0 100 200 300 400 500

Time

14.3 Fourier Transforms and Spectrograms

When working with signals you often have to consider whether or not you
need to work in the time domain or the frequency domain. These two different
domains can provide valuable information about a signal. The most popular
way to explore signals in the frequency domain is to use Fourier Transforms. A
Fourier Transform is a reversible mathematical transform developed by Joseph
Fourier in the early 1800s. The transform breaks apart a time series into a
sum of finite series of sine or cosine functions. This was a huge break through
of it’s time, as Fourier argues that any signal can be broken down into a series
of sine and cosine curves. This was a fantastic development and is still one of
the most widely used tools in digital signal processing, which tells you how
useful it truly is (and how brilliant Fourier was).
The Discrete Fourier Transform (DFT) is a specific form of the Fourier Trans-
form applied to a time wave. The DFT allows you to switch from working
in the time domain to the frequency domain by using the amplitude and the
frequency of the signal to build the frequency spectrum of the original time
wave.
To apply Fourier Transforms we use a window size. The length of the FT
14.3 Fourier Transforms and Spectrograms 299

is controlled by the window length. Increasing the window size will increase
frequency resolution, but this will also make the transform less accurate in
terms of time as more signal will be selected for the transform. Thus there
is a constant level of uncertainty between the time of the signal and the
frequency of the signal2 . Different windows can be used in a DFT depending
on the application. There are numerous windows available within the seewave
package (see Figure 14.3).
1.0

hamming
bartlett
0.8

blackman
flattop
0.6

hanning
rectangle
all

0.4
0.2
0.0

0 100 200 300 400 500

FIGURE 14.3: Different available filters within the seewave package

Computing a fourier transform on a whole sound or a single segment of the

sound may not be enough information for what you desire. One great idea
is to compute the DFT on successive sections along the entire sound signal.
A window is “slided” along the signal and a DFT is computed at each of
these slides. This process is called a Short-Time Fourier Transform. This
method allows you to represent the signals in the frequency spectrum, while
also maintaining most of the information from the time dimension (as Hannah
Montana might say, it’s the best of both worlds3 ).
We can plot the results of an STFT in what is called a spectrogram. The
spectogram is useful in bioacoustics, acoustics, and ecoacoustics (among others)
because it allows you to represent three variables in a 2D plot, which avoids
using a 3D representation that is genereally not appealing. The successive
2 this is a result of the common Heisenberg Uncertainty Principle
3 https://www.youtube.com/watch?v=uVjRe8QXFHY
 300 14 Digital Signal Processing

DFT are plotted against time with the relative amplitude of each sine function
of each DFT represented by some color scale. In other words, the spectrogram
has an x-axis of time, a y-axis of frequency, and the amplitude represented by
the color scheme used inside the plot (which is usually included in a scale).
Let’s use a sample audio file provided by the seewave package and produce a
spectrogram of it using the spectro function. R provides multiple functions
like spectro that make the creation of spectrograms and Fourier Transforms
extremely easy and logical, which is one of the many reasons to perform digital
signal processing within the R environment.

library(seewave)
data(tico)
spectro(tico)

Amplitude
10 (dB)

0
8
Frequency (kHz)

-5

6 -10

-15
4

-20
2
-25

0 -30
0 0.5 1 1.5

Time (s)

Let’s explore more of the parameters for this function to get a better un-
derstanding of how it is working. Type in ?spectro in the R console. The
spectro() function contains a lot of different arguments that allow you to
control how the spectrogram is created. The f argument allows you to specify
the sampling rate of your sound object. The wl option allows you to control
the window length that is used when the successive DFTs are produced (aka
the STFT). You can choose which window you want to use with the wn option.
You can also specify the amount of ovlp between two successive windows with
the ovlp argument. Let’s use some of these arguments and compare it to the
standard spectrogram we previously produced.
 14.3 Fourier Transforms and Spectrograms 301

spectro(tico, f = 22050, wl = 1024, wn = "bartlett", ovlp = 5)

Amplitude
10 (dB)

0
8
Frequency (kHz)

-5

6 -10

-15
4
-20
2
-25

0 -30
0 0.5 1 1.5

Time (s)

You can immediately see that there are some slight differences between the
spectrograms. The first spectrogram appears to be more precise than the second
spectrogram. Explore these parameters yourself to get a better understanding
of how these different properties impact the short-time fourier transform and
the spectrogram.
One cool thing about the spectro() function is that it produces a graph,
but it also stores the values along the three scales. If you save the call to a
spectro() function, that variable will be a list containing the values along
the time axis, values along the frequency axis, and the amplitude values of the
FFT decompositions. These values can be accessed in familiar ways ($time,
$frequency, $amp).
Now that you have a visualization of the sound, we will use the tuneR package to
play the sound so you can see exactly how it matches up with the spectrogram.
First install the tuneR package, which is another package great for working
with sound and signals in R. Then use the play function to play back the
tico sound. Note that the second argument in the play function is the player
argument, which may or may not need to be defined depending on your
operating system and what audio players are installed on your system. In my
case, I am playing the sound through the aplay player on a machine running
Linux.
302 14 Digital Signal Processing

library(tuneR)
play(tico, "aplay")

The audio matches up perfectly with the spectrogram. Pretty cool!

14.4 Analyzing Audio Data with warbleR

Now that we’ve explored how we visualize audio signals with Fourier Transforms
and spectrograms, let’s now explore some of the packages in R used explicitly
for bioacoustic analysis of audio signals. The warbleR package was written to
streamline bioacoustic research inside of R, and it provides numerous different
functions for obtaining, manipulating, and visualizing bioacoustic data (and
especially bioacoustic data of birds). We are going to go through an example
to display some of the many useful functions in the warbleR package. Our
goal is to analyze the recordings of Cedar Waxwings from the xeno-canto
database to determine whether or not there is variation in these calls across
large geographical distances. Along the way we will showcase many useful
functions in the warbleR package, and will compute numerous meausures that
give us useful information about the signal. Let’s get started!
First off, create a new directory where you want to store the data and change
the working directory to that folder. We will produce a lot of files in the
following exercise, so it is best to use a new directory to maintain organization.
Here is the directory we use, you should change yours accordingly:

setwd("/home/jeffdoser/Dropbox/teaching/for875-2020/textbook/")

We will extract sounds from a large and very common database in the field
of bioacoustics, xeno-canto. warbleR allows you to easily query the database
and work with the sounds maintained in their library. We are interested in the
sounds of Cedar Waxwings, which just so happen to be my favorite species of
bird. They are small, beautiful, majestic birds that are fairly common in rural
and suburban areas (Figure 14.4). They also have a very large geographical
region, spanning across all three countries in North America.
The warbleR package allows us to first query the database for the recordings
of Cedar Waxwings (Bombycilla cedrorum) without downloading them.

library(warbleR)
cedarWax <- quer_xc(qword = "Bombycilla cedrorum", download = FALSE)
14.4 Analyzing Audio Data with warbleR 303

FIGURE 14.4: Two cedar waxwings, possibly pondering whether or not Ross
and Rachel will ever get back together (oh wait no that’s me as I write this
chapter between binge watching Friends)

119 recordings found!

names(cedarWax)

[1] "Recording_ID" "Genus"

[3] "Specific_epithet" "Subspecies"
[5] "English_name" "Recordist"
[7] "Country" "Locality"
[9] "Latitude" "Longitude"
[11] "Vocalization_type" "Audio_file"
[13] "License" "Url"
[15] "Quality" "Time"
[17] "Date" "Altitude"
[19] "file.name" "Spectrogram_small"
[21] "Spectrogram_med" "Spectrogram_large"
[23] "Spectrogram_full" "Length"
[25] "Uploaded" "Other_species"
[27] "Remarks" "Bird_seen"
[29] "Playback_used" "Other_species1"
[31] "Other_species2" "Other_species3"
[33] "Other_species4" "Other_species5"
[35] "Other_species6" "Other_species7"
304 14 Digital Signal Processing

[37] "Other_species8"
We use the quer_xc function to query the database, and use the restriction
of the scientific name for Cedar Waxwings so we only obtain recordings of
our desired species. We see that this returns a data frame with a lot of useful
information, such as the latitude and longitude of the recording, the date, the
quality of the recording, and the type of vocalization (since birds have multiple
different types of vocalizations).
Now we produce a map showing the locations of all of the recordings. The
xc_maps function allows you to either save a map as an image to your hard
drive, or you can simply load the image in the R environment. We will load
the image in the R environment by setting the img argument to FALSE.

xc_maps(X = cedarWax, img = FALSE)

The recordings are mostly in the United States, but you can see there are
recordings in Mexico and Canada as well. Now that we have an idea of where
these recordings were taken, let’s look at what types of recordings we have.
We will do this using the table() function.

table(cedarWax$Vocalization_type)

adult, call, sex uncertain

1
adult, sex uncertain, song
1
adult, song
1
alarm call
2
alarm call, call, flight call, juvenile, song
1
begging call, juvenile
3
call
64
Call
2
call, dog
1
call, female, male
1
call, flight call
7
14.4 Analyzing Audio Data with warbleR 305

call, flight call, sex uncertain

1
call, flight call, song
1
call, flock calls
10
call, song
5
call, song, wing sounds
1
Calls of flock
1
Feeding calls
1
flight call
7
Flight call
1
Flight Call
1
juvenile calls
1
Seet, adult, and trill, call, sex uncertain
1
song
4
Look closely at the output from this function. This is a great example of the
messiness of real-world data. Users from around the world input their data
into this website, and this results in many different ways of describing the
vocalizations produced by the birds. You can see the users used many different
ways to describe the calls of cedar waxwings. There are 64 recordings labeled
as “call” so we will just focus on these recordings, but keep in mind that in
a more serious study, we may want to manipulate the labels and combine
vocalizations that are actually of the same type.
Now that we have the recordings we want to work with, we download them
and convert them to wav files. Since this is only an example, we will look
specifically at six of the calls within the cedarWax recordings to minimize
downloading and computation time.4

newCedarWax <- subset(cedarWax, cedarWax$Recording_ID %in% c(313682, 313683, 313684, 32188

quer_xc(X = newCedarWax)

4 Sound files are pretty large so working with these files using the warbleR package can

be time consuming.
306 14 Digital Signal Processing

Downloading files...
double-checking downloaded files
warbleR and seewave are designed to work with wav files while the xeno-canto
database stores its recordings as mp3 files. Luckily there is a very simple
function mp32wav that converts all the mp3 files in the working directory to
wav files. We then remove all the mp3 files from the current directory using
the system function.

mp32wav()
system("rm *.mp3")

warbleR gives us many easy ways to produce spectrograms for all of our desired
sound files at once. To do this, we use the lspec function, which produces
image files with spectrograms of whole sound files split into multiple rows.

lspec(ovlp = 10, it = "tiff", rows = 5)

Now look in your current directory and you should see .tiff files for all of
the sound files. These files could be used to visually inspect sound files and
eliminate undesired files as a result of noise, length, or other variables you are
not interested in. But for this example we will use all six recordings.
Remember that our ultimate goal is to determine whether or not Cedar
Waxwing calls show some sort of variation across geographical distance. We
need to produce some acoustic measures to compare the different recordings.
First, we want to perform signal detection. In other words, we want to find the
exact parts in the recordings where a cedar waxwing is performing its call. We
only want to determine the acoustic parameters during these times, as this will
eliminate noise from having any impact on the parameters. We can perform
signal detection using the autodetec function. This function automatically
detects the start and end of vocalizations based on amplitude, duration, and
frequency range attributes. In addition, autodetec has two types of output:

1. A data frame containing the recording name, selection, start time,

and end time.
2. A spectrogram for each recording with red lines showing the begin-
ning and end of the signal.

autodetec has a TON of parameters that you can use to help specify exactly
what type of signal you are looking for in your data. To help us figure out
what parameters we want to use, let’s first look at an example spectrogram of
one of the recordings (Figure 14.5).
14.4 Analyzing Audio Data with warbleR 307

FIGURE 14.5: Example spectrogram of Cedar Waxwing calls

 308 14 Digital Signal Processing

The calls appear to be between 4-10 kHz in frequency. In this sample they are
quite short, with none lasting longer than a second and most being less than
half a second. Let’s try these two parameters along with a few others and see
what we get. Let’s first do this on just one of the recordings so we don’t waste
computing power if we have to change our parameters

wavs <- list.files(pattern = ".wav", ignore.case = TRUE)

subset <- wavs[2]
autodetec(flist = subset, bp = c(4, 10), threshold = 10, mindur = 0.05,
maxdur = 0.5,ssmooth = 800, ls = TRUE,
res = 100, flim = c(1, 12), wl = 300, set =TRUE, sxrow = 6,
rows = 5, redo = TRUE, it = "jpeg", img =TRUE, smadj = "end")

Detecting signals in sound files:Producing long spectrogram:

sound.files selec start end
1 Bombycilla-cedrorum-313683.wav 1 0.6379 0.8424
2 Bombycilla-cedrorum-313683.wav 2 1.8631 1.9641
3 Bombycilla-cedrorum-313683.wav 3 1.9784 2.1559
4 Bombycilla-cedrorum-313683.wav 4 3.9281 4.3652
5 Bombycilla-cedrorum-313683.wav 5 5.8650 6.2886
6 Bombycilla-cedrorum-313683.wav 6 8.7235 9.0295
7 Bombycilla-cedrorum-313683.wav 7 9.1184 9.4517
Here is a list of the parameters and what each means:
• flist: character vector indicating the subset of files to be analyzed
• bp: numeric vector of length two giving the lower and upper bounds of a
frequency bandpass filter in kHz
• threshold: specifies the amplitude threshold for detecting signals in percent-
age
• mindur, maxdur: specify the minimum call length and maximum call length
respectively
• ssmooth: a numeric vector of length 1 to smooth the amplitude envelope
with a sum smooth function
The rest of the parameters above are controlling the output image. We have
attached the output image below.

knitr::include_graphics("figures/sampleAutodetec.jpg")

Well it looks like we’ve done a pretty good job! There are a couple signals we
are not detecting, but we are not getting any false positives (detecting signals
that aren’t actually signals) so we will continue on with these parameters (feel
free to refine them as much as you want). Let’s run the autodetec function
for all the recordings and save it as a variable. Notice that we switched the
14.4 Analyzing Audio Data with warbleR 309

FIGURE 14.6: Example output of autodetec function

310 14 Digital Signal Processing

img argument to FALSE since we don’t need to produce an image for every
recording. In addition, we remove all null values as this would correspond to
the rare situation in which the autodetec function does not detect a signal in
a sound file

waxwingSignals <- autodetec(flist = wavs, bp = c(4, 10), threshold = 10, mindur = 0.05,
maxdur = 0.5,ssmooth = 800, ls = TRUE,
res = 100, flim = c(1, 12), set =TRUE, sxrow = 6,
rows = 15, redo = TRUE, it = "tiff", img = FALSE, smadj = "end")

Detecting signals in sound files:

waxwingSignals[is.na(waxwingSignals)] <- 0

Now that we have the locations of all the signals in the recordings, we can
begin to compute acoustic parameters. The warbleR package provides the
fantastic specan function that allows us to calculate 22 acoustic parameters
at once through a batch process. This is much more efficient than computing
all these parameters separately. specan uses the time coordinates from the
autodetec function that we saved in our waxwingSignals variable. It then
computes these acoustic parameters for the recordings, but only during the
times when a signal is being performed.

params <- specan(waxwingSignals, bp = c(1, 15))

names(params)

[1] "sound.files" "selec" "duration"

[4] "meanfreq" "sd" "freq.median"
[7] "freq.Q25" "freq.Q75" "freq.IQR"
[10] "time.median" "time.Q25" "time.Q75"
[13] "time.IQR" "skew" "kurt"
[16] "sp.ent" "time.ent" "entropy"
[19] "sfm" "meandom" "mindom"
[22] "maxdom" "dfrange" "modindx"
[25] "startdom" "enddom" "dfslope"
[28] "meanpeakf"
You can see from the names function that there are a wide variety of measures
computed for each sound file. Now that we have all these measurements, let’s
use them to determine whether there is any geographic variation in these cedar
waxwing calls. The first step in doing this is joining the geographic location for
each recording to the params data frame. The geographic location is contained
within the cedarWax data frame. We will first join these tables based on the
recording ID.
14.5 Analyzing Audio Data with warbleR 311

params$Recording_ID <- gsub('.*-([0-9]+).*','\\1', params$sound.files)

newCedarWax$Recording_ID <- as.character(newCedarWax$Recording_ID)
finalData <- inner_join(params, newCedarWax)

Joining, by = "Recording_ID"
We will do this using a process called Principal Components Analysis, a
statistical dimension reduction technique that can be used to find latent
variables that explain large amounts of variation in a data set.

# Remove the non-numeric variables and then run the pca

pca <- prcomp(x = finalData[, sapply(finalData, is.numeric)], scale. = TRUE)
score <- as.data.frame(pca[[5]])
ID <- as.numeric(finalData$Recording_ID)
ggplot(data = score, mapping = aes(PC1, PC2)) +
geom_point(aes(color = finalData$Recording_ID, shape = finalData$Country))

2.5
finalData$Recording_ID
313682
313683
0.0 313684
321881
329907
PC2

361006
-2.5

finalData$Country
Canada
-5.0 Mexico
United States

-5.0 -2.5 0.0 2.5 5.0

PC1

From the graph, we see that two birds from the US form a group with a bird
from Mexico, and a bird from the US forms a group with two birds from
Canada. Depending on where exactly the birds from the US are located, this
could potentially suggest some variation across geographical distance. We leave
this to you to explore further if you so desire. But for now, we see that the
warbleR package has some fantastic tools for manipulating and working with
audio signals within the R environment.
312 14 Digital Signal Processing

14.5 Exercises
Exercise DSP Learning objectives: download, manipulate, and play sound
files using R, produce spectrograms of audio files, use filters, use acoustic
indices to analyze western New York soundscape data}
15
Graphics in R Part 2: graphics

15.1 Scatter Plots

Begin by considering a simple and classic data set sometimes called Fisher’s
Iris Data. These data are available in R.

data(iris)
str(iris)

'data.frame': 150 obs. of 5 variables:

$ Sepal.Length: num 5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
$ Sepal.Width : num 3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
$ Petal.Length: num 1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
$ Petal.Width : num 0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
$ Species : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...
The data contain measurements on petal and sepal length and width for
dim(iris)[1] iris plants. The plants are from one of three species, and the
species information is also included in the data frame. The data are commonly
used to test classification methods, where the goal would be to correctly
determine the species based on the four length and width measurements. To
get a preliminary sense of how this might work, we can draw some scatter
plots of length versus width. In graphics, making scatter plots is simple:

plot(x = iris$Sepal.Length, y = iris$Sepal.Width)

313
314 15 Graphics in R Part 2: graphics

4.0
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

We see the plot() function by default creates a basic scatterplot when we

supply the function with values for the x and y arguments.
Looking at the scatter plot and thinking about the focus of finding a method
to classify the species, two thoughts come to mind. First, the plot might be
improved by shading in the points. And second, using different colors for the
points corresponding to the three species would help.

plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = 19, col = iris$Species)

15.1 Scatter Plots 315

4.0
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

The pch = argument is used to control the type of symbol used to represent the
data points. pch = 19 specifies a filled in circle for each data point. You can
mess around with different numbers and you’ll see a wide variety of options. The
col = iris$Species argument tells the plot() function to assign a different
color to each unique value in the the iris$Species vector.1 Notice however
that there is no legend produced by default, so we don’t know which color
represents each species. We can add a legend using the legend() function.

plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = 19, col = iris$Species)

legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = 19)

1 To see the order in which R assigns colors to different values, run the palette() function
316 15 Graphics in R Part 2: graphics

setosa
versicolor
4.0

virginica
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

Let’s walk through the additional complexities one at a time:

1. x = 6.4, y = 4.4 specifies the location where the legend is drawn.

Here I place it in a location where the box will not cover up any
data points.
2. legend = levels(iris$Species) specifies the names of the differ-
ent categories that will appear in the legend.
3. col = 1:length(levels(iris$Species)) specifies the number of
colors that are represented in the function. In this case, we use
length(levels(iris$Species)) to get the number of unique
species in the data set, which turns out to be 3.
4. pch = 19 specifies the type of symbol to include in the legend.

It may seem like there are a lot of arguments to specify to produce a legend
and a simple scatter plot, but this also means we have a lot of control over the
graph, which is very useful when you are trying to develop a publication-quality
figure for your own work.
Now perhaps we want to use different shapes as well for the different species.
To do this we assign different values for the pch argument to each species, and
subsequently change the pch argument in the legend as well.

plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = c(17, 18, 19)[iris$Species], col =

legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = c(17, 18, 19))
15.1 Scatter Plots 317

setosa
versicolor
4.0

virginica
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

15.1.1 Adding lines to a scatter plot

Next let’s add a fitted least squares line to the scatter plot. This can be done
pretty easily using the abline() function to add a straight ine to the curve,
and the lm() function to actually compute the least squares line

plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = c(17, 18, 19)[iris$Species], col =

legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = c(17, 18, 19))
abline(lm(iris$Sepal.Width ~ iris$Sepal.Length))
318 15 Graphics in R Part 2: graphics

setosa
versicolor
4.0

virginica
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

For the iris data, it probably makes more sense to fit separate lines by species.
Below we compute the linear model separately for each species, then use the
abline() function to add each line to the graph. Note our use of logical
subsetting to obtain separate data frames for each species.

iris.setosa <- iris[iris$Species == "setosa", ]

iris.versicolor <- iris[iris$Species == "versicolor", ]
iris.virginica <- iris[iris$Species == "virginica", ]
plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = c(17, 18, 19)[iris$Species], col =
legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = c(17, 18, 19))
abline(lm(iris.setosa$Sepal.Width ~ iris.setosa$Sepal.Length), col = 1)
abline(lm(iris.versicolor$Sepal.Width ~ iris.versicolor$Sepal.Length), col = 2)
abline(lm(iris.virginica$Sepal.Width ~ iris.virginica$Sepal.Length), col = 3)
15.1 Scatter Plots 319

setosa
versicolor
4.0

virginica
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

Note that we could also do this using a for loop, which you will learn about
soon in Chapter 7.

iris.setosa <- iris[iris$Species == "setosa", ]

iris.versicolor <- iris[iris$Species == "versicolor", ]
iris.virginica <- iris[iris$Species == "virginica", ]
plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = c(17, 18, 19)[iris$Species], col =
legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = c(17, 18, 19))
for (i in 1:length(levels(iris$Species))) {
curr.species <- iris[iris$Species == levels(iris$Species)[i], ]
abline(lm(curr.species$Sepal.Width ~ curr.species$Sepal.Length), col = i)
}
320 15 Graphics in R Part 2: graphics

setosa
versicolor
4.0

virginica
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

Another common use of line segments in a graphic is to connect the points

in order. Although it is not clear why this helps in understanding the iris
data, the technique is illustrated next. First we connect all lines regardless
of species. This is done easily by specifying the argument type = 'b' in the
plot() function2

plot(x = iris$Sepal.Length, y = iris$Sepal.Width, pch = c(17, 18, 19)[iris$Species], col =

legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = c(17, 18, 19))

2 The “b” stands for “both”. By default, the type argument is set to “p” for “points”. You

can also specify that you only want lines using type = 'l'.
15.1 Scatter Plots 321

setosa
versicolor
4.0

virginica
iris$Sepal.Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

iris$Sepal.Length

Now we attempt to only connect the points within a given species. Doing this
requires the use of the lines() function to draw three separate lines for each
species3 First we create a plot for the first species, and then subsequently
use the lines() function to add each of the remaining two species to the
plot. Notice that there are some additional arguments in the original plot()
function that we have yet to cover. Don’t worry, we’ll talk about all this fun
stuff in the next section :)

plot(x = iris$Sepal.Length[iris$Species == "setosa"], y = iris$Sepal.Width[iris$Species ==

lines(x = iris$Sepal.Length[iris$Species == "versicolor"], y = iris$Sepal.Width[iris$Speci
lines(x = iris$Sepal.Length[iris$Species == "virginica"], y = iris$Sepal.Width[iris$Specie
legend(x = 6.4, y = 4.4, legend = levels(iris$Species),
col = 1:length(levels(iris$Species)), pch = c(17, 18, 19))

3 Note that we could again perhaps simplify this by using a for loop, but we’ll restrain

ourselves until we learn about for loops in Chapter 7.

322 15 Graphics in R Part 2: graphics

setosa
versicolor
4.0

virginica
Sepal Width

3.5
3.0
2.5
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

Sepal Length

15.2 Labels, Axes, Text, etc.

The default settings of graphics often produce useful graphics, but once a
graphic is chosen for dissemination, the user will likely want to customize things
like the title, axes, etc. In this section some tools for customization are presented.
Most will be illustrated in the context of a data set on crime rates in the 50 states
in the United States. These data were made available by Nathan Yau at http:
//flowingdata.com/2010/11/23/how-to-make-bubble-charts/. The data
include crime rates per 100,000 people for various crimes such as murder and
robbery, and also include each state’s population. The crime rates are from
the year 2005, while the population numbers are from the year 2008, but
the difference in population between the years is not great, and the exact
population is not particularly important for what we’ll do below.
First, read in the data, examine its structure, and produce a simple scatter
plot of motor vehicle theft versus burglary.

u.crime <- "https://www.finley-lab.com/files/data/crimeRatesByState2005.csv"

crime <- read.csv(u.crime, header=TRUE)
str(crime)
 15.2 Labels, Axes, Text, etc. 323

'data.frame': 50 obs. of 9 variables:

$ state : chr "Alabama " "Alaska " "Arizona " "Arkansas" ...
$ murder : num 8.2 4.8 7.5 6.7 6.9 3.7 2.9 4.4 5 6.2 ...
$ Forcible_rate : num 34.3 81.1 33.8 42.9 26 43.4 20 44.7 37.1 23.6 ...
$ Robbery : num 141.4 80.9 144.4 91.1 176.1 ...
$ aggravated_assult : num 248 465 327 387 317 ...
$ burglary : num 954 622 948 1085 693 ...
$ larceny_theft : num 2650 2599 2965 2711 1916 ...
$ motor_vehicle_theft: num 288 391 924 262 713 ...
$ population : int 4627851 686293 6500180 2855390 36756666 4861515 3501252 87309

plot(crime$burglary, crime$motor_vehicle_theft, pch = 19)

800 1000
crime$motor_vehicle_theft

600
400
200

400 600 800 1000 1200

crime$burglary

15.2.1 Labels

By default, axis labels are the values of the x and y values you supply to the
plot() function. For making a publication-quality figure, we often want to
further customize these labels, as well as potentially add a title to the graph.
We can do this using the xlab, ylab, and main arguments in the plot()
function.
 324 15 Graphics in R Part 2: graphics

plot(crime$burglary, crime$motor_vehicle_theft, pch = 19,

xlab = "Burglaries per 100,000 population",
ylab = "Motor vehicle theft per 100,000 population",
main = "Burglaries vs motor vehicle theft for US states")

Burglaries vs motor vehicle theft for US states

Motor vehicle theft per 100,000 population

800 1000
600
400
200

400 600 800 1000 1200

Burglaries per 100,000 population

Perhaps it may be easier to read the values on the y-axis if they were oritented
horizontally instead of vertically. We can do this using the las = 1 argument.

plot(crime$burglary, crime$motor_vehicle_theft, pch = 19,

xlab = "Burglaries per 100,000 population",
ylab = "Motor vehicle theft per 100,000 population",
main = "Burglaries vs motor vehicle theft for US states",
las = 1)
 15.3 Customizing Axes 325

Burglaries vs motor vehicle theft for US states

Motor vehicle theft per 100,000 population

1000

800

600

400

200

400 600 800 1000 1200

Burglaries per 100,000 population

15.3 Customizing Axes

The basic plot() function provides defaults for the axis limits, as well as
for other axis features (i.e., the box around the plot). These, and other axis
features such as tick marks, labels, and transformations, can all be customized.
First, we use the xlim and ylim arguments to make the x and y axes start at
zero and go to the maxium x and y values.

plot(crime$burglary, crime$motor_vehicle_theft, pch = 19,

xlab = "Burglaries per 100,000 population",
ylab = "Motor vehicle theft per 100,000 population",
main = "Burglaries vs motor vehicle theft for US states",
las = 1, xlim = c(0, max(crime$burglary)),
ylim = c(0, max(crime$motor_vehicle_theft)))
 326 15 Graphics in R Part 2: graphics

Burglaries vs motor vehicle theft for US states

Motor vehicle theft per 100,000 population

1000

800

600

400

200

0 200 400 600 800 1000 1200

Burglaries per 100,000 population

Now we eliminate the axes and box automatically drawn by the plot()
function, and subsequently add our own axes using the axis() function. We
use most of the defaults provided by the axis() function, however, looking at
the help page for axis() reveals a large amount of possibilities for customizing
the axes in whatever way you desire.

plot(crime$burglary, crime$motor_vehicle_theft, pch = 19,

xlab = "Burglaries per 100,000 population",
ylab = "Motor vehicle theft per 100,000 population",
main = "Burglaries vs motor vehicle theft for US states",
las = 1, axes = FALSE,
xlim = c(0, max(crime$burglary)),
ylim = c(0, max(crime$motor_vehicle_theft)))
axis(side = 1)
axis(side = 2, las = 1)
 15.3 Customizing Axes 327

Burglaries vs motor vehicle theft for US states

Motor vehicle theft per 100,000 population

1000

800

600

400

200

0 200 400 600 800 1000 1200

Burglaries per 100,000 population

15.3.1 Text, Point Size, and Color

Next we make point size proportional to population, change the color, and add
a state label. First we make point size proportional to population size using
the symbols() function. We can control the color of the points using the bg
argument in the symbols() function. We then use the text() function to add
labels to each point.

symbols(crime$burglary, crime$motor_vehicle_theft, pch = 19,

xlab = "Burglaries per 100,000 population",
ylab = "Motor vehicle theft per 100,000 population",
main = "Burglaries vs motor vehicle theft for US states",
las = 1,
xlim = c(0, max(crime$burglary)),
ylim = c(0, max(crime$motor_vehicle_theft)),
circles = crime$population,
inches = 1/5, axes = FALSE, bg="steelblue2")
axis(side = 1)
axis(side = 2, las = 1)
text(crime$burglary, crime$motor_vehicle_theft, labels=crime$state, cex=0.9, font=2)
 328 15 Graphics in R Part 2: graphics

Burglaries vs motor vehicle theft for US states

Motor vehicle theft per 100,000 population

Nevada
1000
Arizona
800 Washington
Hawaii
California
600 Maryland
Colorado
Oregon
Michigan Georgia
Rhode Island Missouri Tennessee
Florida
TexasNew Mexico
400 Alaska Oklahoma
South
Ohio Carolina
New Jersey Utah
Nebraska Indiana Louisiana
Kansas North Carolina
Illinois
Massachusetts
Connecticut Alabama
Delaware Mississippi
Minnesota
Pennsylvania Arkansas
200 Wisconsin
Virginia
Montana Kentucky
West
Idaho Virginia
New
North York
Dakota Iowa
Wyoming
New Hampshire
South Dakota
Vermont
Maine
0

0 200 400 600 800 1000 1200

Burglaries per 100,000 population

This graph is mind-numbingly busy, and is missing some notable things (i.e. leg-
end). For now, we leave it up to you to add in a legend, make the graph less
cluttered, fix the cutoff labels, etc. if you so desire.
You run an experiment to see if the number of alcoholic beverages a person
has on average every day is related to weight. You collect 15 data points. Enter
these data in R by creating vectors, and then reproduce the following plot.

# of Drinks Weight
1.0 150
3.0 350
2.0 200
0.5 140
2.0 200
1.0 160
0.0 175
0.0 140
0.0 167
1.0 200
4.0 300
5.0 321
2.0 250
0.5 187
1.0 190
15.4 Other Types of Graphics 329

350

300
Weight

250

200

150

0 1 2 3 4 5

Number of Drinks

15.4 Other Types of Graphics

Scatter and line plots, which have just been presented, are common but
certainly not the only graphical displays in common use. Histograms, boxplots,
and bar graphs, as well as more “mathematical” displays such as the graph
of a function, are commonly used to represent data. Examples of each are
presented below.

15.4.1 Histograms

Simon Newcomb conducted several experiments to estimate the speed of light

by measuring the time it took for light to travel from his laboratory to a mirror
at the base of the Washington Monument, and then back to his lab. This is a
distance of 7.44373 km, and by dividing this distance by the measured time,
an estimate for the speed of light is obtained.
The times are of course quite small, and to avoid working with very small
numbers, the data are recoded to be the deviation from 24800 nanoseconds. For
example an observation coded as 28 represents a time of 24828 nanoseconds,
while an observation coded as −44 represents a time of 24756 nanoseconds.
330 15 Graphics in R Part 2: graphics

u.newcomb <- "https://www.finley-lab.com/files/data/Newcomb.csv"

Newcomb <- read.csv(u.newcomb, header=TRUE)
head(Newcomb)

Time
1 28
2 26
3 33
4 24
5 34
6 -44
To produce a simple histogram of these data in the graphics package, we can
use the simple hist() function.

hist(Newcomb$Time)

Histogram of Newcomb$Time
40
30
Frequency

20
10
0

-40 -20 0 20 40

Newcomb$Time

The function automatically specifies the binwidths for the data, however we
can easily change this ourselves to something more suitable using the breaks
argument. We also change the axis labels, y-axis label orientation, title, and
color of the bins.
15.4 Other Types of Graphics 331

hist(Newcomb$Time, breaks = 30, main = "Histogram", xlab = "Time", las = 1, col = 'blue')

Histogram

12

10

8
Frequency

-40 -20 0 20 40

Time

15.4.2 Boxplots

Next we consider some data from the gap minder data set to construct some
box plots. These data are available in the gapminder package, which might
need to be installed via install.packages("gapminder").

library(gapminder)

Attaching package: 'gapminder'

The following object is masked _by_ '.GlobalEnv':

gapminder

boxplot(gdpPercap ~ continent, data = subset(gapminder, year == 2002), pch = 19, col = 'li
 332 15 Graphics in R Part 2: graphics

40000
GDP Per-capita

30000

20000

10000

Africa Americas Asia Europe Oceania

Continent

Here’s the same set of boxplots, but with different colors and the boxes plotted
horizontally rather than vertically

boxplot(gdpPercap ~ continent, data = subset(gapminder, year == 2002), pch = 19, border =

Oceania

Europe
Continent

Asia

Americas

Africa

0 10000 20000 30000 40000

GDP Per-Capita
 15.4 Other Types of Graphics 333

However, notice in the last two graphs the y-axis labels were running off
the page. Using the graphics package, we have the ability to adjust all the
dimensions and parameters of a plot by using the par function. Below we
utilize the par function to extend the amount of space in the plot on the left
side of the graph.

par(mar=c(6,4,4,2))
boxplot(gdpPercap ~ continent, data = subset(gapminder, year == 2002), pch = 19, border =

Oceania

Europe
Continent

Asia

Americas

Africa

0 10000 20000 30000 40000

GDP Per-Capita

15.4.3 Bar Graphs

As part of a study, elementary school students were asked which was more
important to them: good grades, popularity, or athletic ability. Here is a brief
look at the data.

u.goals <- "https://www.finley-lab.com/files/data/StudentGoals.csv"

StudentGoals <- read.csv(u.goals, header=TRUE)
head(StudentGoals)

Gender Grade Age Race Type School Goals Grades

1 boy 5 11 White Rural Elm Sports 1
2 boy 5 10 White Rural Elm Popular 2
3 girl 5 11 White Rural Elm Popular 4
334 15 Graphics in R Part 2: graphics

4 girl 5 11 White Rural Elm Popular 2

5 girl 5 10 White Rural Elm Popular 4
6 girl 5 11 White Rural Elm Popular 4
Sports Looks Money
1 2 4 3
2 1 4 3
3 3 1 2
4 3 4 1
5 2 1 3
6 2 1 3
First, a simple bar graph of the most important goal chosen is drawn, followed
by a stacked bar graph which also includes the student’s gender. We then add
a side by side bar graph that includes the student’s gender.

counts <- table(StudentGoals$Goals)

counts

Grades Popular Sports

247 141 90

barplot(counts)
200
150
100
50
0

Grades Popular Sports

Notice that to create a basic bar graph using the graphics package, we first
take our variable of interest StudentGoals$Goals and put it in table format
15.4 Other Types of Graphics 335

using the table() function. Next we produce a stacked bar graph that also
includes the student’s gender. Then we subsequently add a side by side bar
graph that includes the student’s gender.

gender.counts <- table(StudentGoals$Gender, StudentGoals$Goals)

barplot(gender.counts, col = c("red", "lightblue"), las = 1, ylab = "Count", xlab = "Goals
legend(x = "topright", legend = row.names(gender.counts), col = c("red", "lightblue"), pch

boy
girl
200

150
Count

100

50

0
Grades Popular Sports

Goals

gender.counts <- table(StudentGoals$Gender, StudentGoals$Goals)

barplot(gender.counts, col = c("red", "lightblue"), las = 1, ylab = "Count", xlab = "Goals
legend(x = "topright", legend = row.names(gender.counts), col = c("red", "lightblue"), pch
336 15 Graphics in R Part 2: graphics

120 boy
girl
100

80
Count

60

40

20

0
Grades Popular Sports

Goals

In this example R counted the number of students who had each goal and
used these counts as the height of the bars. Sometimes the data contain the
bar heights as a variable. For example, we create a bar graph of India’s per
capita GDP with separate bars for each year in the data4 .

india <- subset(gapminder, country == "India")

gdp.india <- india$gdpPercap
names(gdp.india) <- india$year
barplot(gdp.india, xlab = "Year", ylab = "GDP Per Capita", main = "India's per-capita GDP"

4R offers a large color palette, run colors() on the console to see a list of color names.
15.4 Other Types of Graphics 337

India's per-capita GDP

1000 1500 2000
GDP Per Capita

500
0

1952 1962 1972 1982 1992 2002

Year

15.4.4 Graphs of Functions

One way to create a plot of a mathematical function f is to create a data

frame with x values in one column and f (x) values in another column, and
then draw a line plot.

x <- seq(-pi, pi, len = 1000)

sin.x <- sin(x)
plot(x, sin.x, type = 'l')
338 15 Graphics in R Part 2: graphics

1.0
0.5
sin.x

0.0
-0.5
-1.0

-3 -2 -1 0 1 2 3

15.5 Saving Graphics

We often want to export our graphics to use in an external document or share
with colleagues. There are numerous functions available to export graphic files
in different formats. Below we demonstrate the use of the png() function to
export the image of the sin curve we created above.

png("sinCurve.png")
plot(x, sin.x, type = 'l')
dev.off()

Inside the png() function, we specify the name the image is given when it is
saved. There are also arguments to specify the width and height of the saved
image. This function opens a png file where the image will be saved. There are
analagous functions for other types of files (i.e. pdf(), jpeg()). The second
line plot(x, sin.x, type = 'l') simply creates the plot we desire to save.
The dev.off() function closes the png file.
15.6 A Summary of Useful graphics Functions and Arguments 339

15.6 A Summary of Useful graphics Functions and Ar-

guments
Upon initial learning, making plots with the graphics package may seem
overwhelming due to the variety of different functions and arguments that
control different features of the image that is produced. However, after some
practice, you will find that the graphics method of graphing gives you great
flexibility in making your images just the way you like them. This flexibility
and control is one of the main reasons we produce most of our visualizations
using graphics.
Below we provide a summary of the major functions and arguments you may
find helpful when working with the graphics package. This is certainly not an
exhaustive list, but it does include functions and arguments we often use on a
daily basis, so we think you’ll find it useful. One of the nice things about the
graphics package is that many of these functions utilize the same arguments.

1. plot(x, y): the basic plotting function. Many of the graphs you
produce will start with the plot() function. Here are some useful
arguments (aside from x and y arguments to specify the data):
•pch: specify the type of point in a scatter plot. We often use
pch = 19.
•xlab and ylab: specify the x and y axis labels
•xlim and ylim: specify the range of the x and y axes
•type: specify the type of the plot. By default it is p, can create
a line plot with type = 'l' or a scatter plot with lines with
type = 'b'.
•main: add a title to the plot
•cex, cex.lab, cex.main: change the sizes of the points, the size
of the axis label text, and the size of the title text, respectively
•col: specify the color of the points/lines
•axes: if TRUE axes will be automatically drawn, if FALSE no
axes will be drawn.
2. legend(x, y, legend): add a legend to the graph specified by the
location (x, y) with the text specified by the legend argument.
3. points(x, y): add points to an already existing plot. Arguments
same as those of plot().
4. lines(x, y): add lines to an already existing plot. Arguments same
as those of plot().
5. axis(): specify the axes
6. text(x, y): add text to a plot at the location (x, y).
7. par(): function to change a wide variety of graphical parameters
for finer control over the plot’s appearance.
340 15 Graphics in R Part 2: graphics

The functions listed above certainly do not cover the realm of plots you can
produce using the graphics package, but having a good understanding of how
to use these functions will give you a firm grip on making publication-quality
plots in R.
For additional help, we find that the help pages for most graphics functions
are quite helpful. The online documentation for the package available here5
perhaps provides a more user friendly version of the help pages within R. There
are also numerous books and online tutorials available for working with these
functions, so a simple google search is often the best way to find additional
help :)

15.6.1 Practice Problem

Go to the help page for the par() function (use help(par) or ?par()). Look
at the plethora of different arguments that you can use to specialize your plot
just the way you like it. List 5 arguments that we didn’t use in this chapter,
describe what they do, then use them in a graph using the iris data set (you
can use any of the variables in the data set and produce any type of graph
you like).

5 https://rdrr.io/r/graphics/graphics-package.html#heading-0
16
Solutions to Practice Problems

2.3.5

a <- (exp(14) + log10(8)) * sqrt(5)

b <- log(4) - (5 * 10ˆ2)
a / b

4.1.3

tree.sp[length(tree.sp) - 1]

4.4.1

my.weights[my.weights == 12] <- NA

sd(my.weights, na.rm = TRUE)

341
342 16 Solutions to Practice Problems

4.8.1

sim.vals <- rnorm(10000)

hist(sim.vals)

5.3.2

ggplot(data = alc.dat, aes(x = alc.drinks, y = alc.weights)) +

geom_point(col = 'blue', size = 2, shape = "square") +
labs(x = "Number of Drinks",
y = "Weight")

6.3.3 (a)

apply(WeatherKLAN2014Full[,c(2:19, 21, 23)], 2, sd, na.rm = TRUE)

6.3.3 (b)

apply(iris[, -5], 2, median)

16.0 343

6.6.2

separate(data = data.birds, col = recordingInfo, sep = "-",

into = c("site", "year", "month", "day"))

6.7.4

select(gm, contains('c'))

6.7.7

gm %>%
filter(country == 'Afghanistan') %>%
select(c("year", "lifeExp")) %>%
arrange(desc(lifeExp))

6.7.11

iris %>%
mutate(s.p.ratio = Sepal.Length / Petal.Length) %>%
group_by(Species) %>%
summarize(mean.ratio = mean(s.p.ratio)) %>%
arrange(desc(mean.ratio))
344 16 Solutions to Practice Problems

7.1.1

FtoK <- function(F) {

K <- ((F - 32) * (5 / 9)) + 273.15
return(K)
}

7.4.4

a[a > 1] <- 0

10.2.2

falsePositives <- rep(0, 40)

for (i in 1:length(falsePositives)) {
knn_Pima <- knn(Pima.tr[,c(2,5)], Pima.te[,c(2,5)], Pima.tr[,8], k = i, prob=TRUE)
falsePositives[i] <- table(knn_Pima, Pima.te[,8])[2, 1]
}
# Using graphics
plot(falsePositives, pch = 19, las = 1, xlab = "k", ylab = "# of False Positives")
# Using ggplot2
dat <- data.frame(k = 1:40, falsePositives)
ggplot(data = dat, aes(x = k, y = falsePositives)) +
geom_point() +
labs(y = "# of False Positives")
16.0 345

11.3.3

unlist(strsplit(strings, split = '[-|.]'))

11.3.5

library(wordcloud)
md.data.frame <- as.data.frame(moby_dick_word_table)
wordcloud(md.data.frame$moby_dick, md.data.frame$Freq, max.words = 500, colors = rainbow(2
Bibliography

Bivand, R. S., Pebesma, E., and Gomez-Rubio, V. (2013). Applied spatial data
analysis with R, Second edition. Springer, NY.
Hastie, T., Tibshirani, R., and Friedman, J. (2009). The elements of statistical
learning: data mining, inference and prediction. Springer, 2 edition.
James, G., Witten, D., Hastie, T., and Tibshirani, R. (2014). An Introduction
to Statistical Learning: with Applications in R. Springer Texts in Statistics.
Springer New York.
Spector, P. (2008). Data Manipulation With R. Use R!
Wickham, H. and Sievert, C. (2016). ggplot2: Elegant Graphics for Data
Analysis. Springer, Cham, 2nd edition.

347