Scientific data visualization

Using ggplot2

Sacha Epskamp

University of Amsterdam
Department of Psychological Methods

11-04-2014
Hadley Wickham
Hadley Wickham
Evolution of data visualization
Scientific data visualization

I Data and analysis results are best communicated through

visualizations
I The leading software for statistical analyses is the
statistical programming language R
I The leading R extension for data visualization is ggplot2
I This presentation will quickly teach you strong visualization
techniques in R
First use of R

I We will use the environment RStudio for our work in R

I RStudio has 4 panels:
Console This is the actual R window, you can enter
commands here and execute them by
pressing enter
Source This is where we can edit scripts. It is where
you should always be working. Control-enter
sends selected codes to the console
Plots/Help This is where plots and help pages will be
shown
Workspace Shows which objects you currently have
I Anything following a # symbol is treated as a comment!
R workflow

I File → New File → R script

I Write codes in the R script
I Select codes and press control + enter to execute
them
Import data

File <- "http://sachaepskamp.com/files/OPdata.csv"

Data <- read.csv(File)
Look at data
head(Data)

## userID Measurement Gender Age Study Work Neuroticism

## 1 1 1 female 24 yes part time low
## 2 1 2 female 24 yes part time low
## 3 1 3 female 24 yes part time low
## 4 1 4 female 24 yes part time low
## 5 1 5 female 24 yes part time low
## 6 1 6 female 24 yes part time low
## Extraversion Openness Conscienciousness Agreeableness
## 1 low high high high
## 2 low high high high
## 3 low high high high
## 4 low high high high
## 5 low high high high
## 6 low high high high
## Stress
## 1 0.375
## 2 0.875
## 3 1.375
## 4 1.875
## 5 1.875
## 6 0.750
Look at data

names(Data)

## [1] "userID" "Measurement"

## [3] "Gender" "Age"
## [5] "Study" "Work"
## [7] "Neuroticism" "Extraversion"
## [9] "Openness" "Conscienciousness"
## [11] "Agreeableness" "Stress"
Look at data

str(Data)

## 'data.frame': 750 obs. of 12 variables:

## $ userID : int 1 1 1 1 1 1 1 1 1 1 ...
## $ Measurement : int 1 2 3 4 5 6 7 8 9 10 ...
## $ Gender : Factor w/ 2 levels "female","male": 1 1 1 1 1
## $ Age : int 24 24 24 24 24 24 24 24 24 24 ...
## $ Study : Factor w/ 2 levels "no","yes": 2 2 2 2 2 2 2
## $ Work : Factor w/ 3 levels "full time","none",..: 3 3
## $ Neuroticism : Factor w/ 2 levels "high","low": 2 2 2 2 2 2
## $ Extraversion : Factor w/ 2 levels "high","low": 2 2 2 2 2 2
## $ Openness : Factor w/ 2 levels "high","low": 1 1 1 1 1 1
## $ Conscienciousness: Factor w/ 2 levels "high","low": 1 1 1 1 1 1
## $ Agreeableness : Factor w/ 2 levels "high","low": 1 1 1 1 1 1
## $ Stress : num 0.375 0.875 1.375 1.875 1.875 ...
Look at data

View(Data)
ggplot2

I ggplot2 (Wickham, 2009) is an implementation of the

Grammer of Graphics (Wilkinson, Wills, Rope, Norton, &
Dubbs, 2006)
I Very different from base R plotting but also very flexible
and powerfull
I Uses data frames as input
I Data must be in long format
I This means that each row is an observation and each
column a variable
I Use reshape2 to get data in long format
I Also check out dplyr (http://sachaepskamp.com/
files/dplyrTutorial.html)
Basics of a plot

I A plot is a 2D repressentation of data, in which variables

can be visualized by, e.g.,:
I Horizontal placing
I Vertical placing
I Color
I Different Lines
I Line type
I Size
I Shape
I ...
I These are called aesthetics
I In ggplot2 we first set aesthetic mapping of our data
using aes() inside ggplot()
I Which variables will be mapped to which aesthetics?
install.packages("ggplot2")
library("ggplot2")
ggplot(Data, aes(x = Measurement, y = Stress))

## Error: No layers in plot

Geometrics

I Next, we define how these aesthetics are used and what

we are plotting:
I Lines
I Points
I Boxplots
I Curves
I ...
I These are called geometrics (geoms)
I We can add these to the plot using +
ggplot(Data, aes(x = Measurement, y = Stress)) +
geom_point()

2
Stress

0 10 20 30
Measurement
ggplot(Data, aes(x = Measurement, y = Stress)) +
geom_boxplot()

2
Stress

10 20
Measurement
ggplot(Data, aes(x = Measurement, y = Stress, group = Measurement)) +
geom_boxplot()

2
Stress

0 10 20 30
Measurement
ggplot(Data,
aes(x = Measurement, y = Stress, group = userID)
) + geom_line()

2
Stress

0 10 20 30
Measurement
ggplot(Data,
aes(x = Measurement, y = Stress, group = userID,
colour = Age)) + geom_line() +
facet_grid(Gender ~ .)

female
1

Age

50
0
Stress

40
3
30

20

male
1

0
0 10 20 30
Measurement
Store elements in an object:
g <- ggplot(Data,
aes(x = Measurement, y = Stress, group = userID,
colour = Age))
g <- g + geom_line()
g <- g +facet_grid(Gender ~ .)
Print the object to plot:
print(g)

female
1

Age

50
0
Stress

40
3
30

20

male
1

0
0 10 20 30
Measurement
I Many more graphical options can be added to ggplot calls
xlab Label of x-axis
ylab Label of y -axis
ggtitle Title of plot
theme Many, many graphical settings
theme_bw() A default black and white theme
I Use Google!
g + xlab("Time") + ylab("Amount of Stress") +
ggtitle("A very fancy plot") + theme_bw()

A very fancy plot

3

female
1
Age
Amount of Stress

50
0
40
3
30

20
2

male
1

0
0 10 20 30
Time
str(sumData)

## 'data.frame': 66095 obs. of 5 variables:

## $ user_id : num 1456 1713 1837 1845 21167 ...
## $ rating : num 0.482 -5.225 -6.639 -0.417 -3.008 ...
## $ date_of_birth: Date, format: "2001-06-03" ...
## $ grade : num 8 8 8 8 7 8 7 8 8 8 ...
## $ gender : chr "f" "m" "m" "f" ...
ggplot(sumData, aes(x = date_of_birth, y = rating)) +
geom_point()

20

10

0
rating

-10

-20

2002 2004 2006 2008 2010

date_of_birth
ggplot(sumData, aes(x = date_of_birth, y = rating)) +
stat_binhex()

20

10

count
800
0
600
rating

400

200

-10

-20

2002 2004 2006 2008 2010

date_of_birth
ggplot(sumData, aes(x = date_of_birth, y = rating,
colour = factor(grade))) + geom_point()

20

10
factor(grade)
1
2
3
0
rating

4
5
6
7
-10
8

-20

2002 2004 2006 2008 2010

date_of_birth
ggplot(sumData, aes(x = date_of_birth, y = rating,
colour = factor(grade), fill = factor(grade))) +
geom_point() + geom_smooth(col = "black", method = "lm")

20

10
factor(grade)
1
2
3
0
rating

4
5
6
7
-10
8

-20

2002 2004 2006 2008 2010

date_of_birth
ggplot(sumData, aes(x = date_of_birth, y = rating,
colour = factor(grade), fill = factor(grade))) +
geom_point() + geom_smooth(col = "black", method = "lm",
formula = y ~ poly(x, 2))

20

10
factor(grade)
1
2
3
0
rating

4
5
6
7
-10
8

-20

2002 2004 2006 2008 2010

date_of_birth
ggplot(sumData, aes(x = grade)) + geom_histogram()

12000

9000
count

6000

3000

2 4 6 8
grade
ggplot(sumData,
aes(x = grade, y = rating, colour = factor(grade))
) + geom_violin()

20

10
factor(grade)
1
2
3
0
rating

4
5
6
7
-10
8

-20

2 4 6 8
grade
 Betweenness Closeness Strength Zhang Onnela
Xss
Xso
Xsb
Xli
Oun
Oin
Ocr
Oaa
Hsi
Hmo
Hga
Hfa
Ese
Efe
Ede
Ean
Cpr
Cpe
Cor
Cdi
Apa
Age
Afo
Afl

0 10 20 30 0.0020 0.0025 0.0030 0.0035 0.6 0.9 1.2 1.5 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15
 Spel: mollenspel
20

Sequence
length
1
2
10
3
Rating

4
5
6
7
0 8
9

Apr 01 Apr 15 May 01 May 15 Jun 01

Date
 Spel: mollenspel

20

Sequence
length
1
2
10 3
Rating

4
5
6
7
0
8
9

−10

Nov 03 Nov 10 Nov 17 Nov 24 Dec 01 Dec 08 Dec 15

Date
 woord.pdf
8

4
userRating

−4

Nov 15 Dec 01 Dec 15 Jan 01 Jan 15 Feb 01

created
 woord.pdf
Binned by item
8

Number of items made

500
Rating

400
300
0
200
100

−4

Nov 15 Dec 01 Dec 15 Jan 01 Jan 15 Feb 01

 woord.pdf
Binned by item
8

Score − expected
0.5
Rating

0.0
0
−0.5

−1.0

−4

Nov 15 Dec 01 Dec 15 Jan 01 Jan 15 Feb 01

 woord.pdf
Binned by item
8

Response time
40000
Rating

30000

0 20000

10000

−4

Nov 15 Dec 01 Dec 15 Jan 01 Jan 15 Feb 01

More ggplot2
More ggplot2
More ggplot2
likert package (Bryer & Speerschneider, 2013)
More ggplot2
sjPlot package (Lüdecke, 2014)
More ggplot2
sjPlot package (Lüdecke, 2014)
More ggplot2
sjPlot package (Lüdecke, 2014)
More ggplot2
sjPlot package (Lüdecke, 2014)
ggplot2 conclusion

I ggplot2 can create very complex visualizations with

minimal codes
I Automatizes convenient things such as
I Margins
I Legend
I Documentation: http://docs.ggplot2.org/
 Thank you for your attention!
Exercises are on http://sachaepskamp.com/files/
ggplot2_exercises.html
References I

Bryer, J., & Speerschneider, K. (2013). likert: Functions to

analyze and visualize likert type items [Computer
software manual]. Retrieved from
http://CRAN.R-project.org/package=likert
(R package version 1.1)
Lüdecke, D. (2014). sjplot: sjplot - data visualization for
statistics in social science [Computer software manual].
Retrieved from
http://CRAN.R-project.org/package=sjPlot
(R package version 1.3)
Wickham, H. (2009). ggplot2: elegant graphics for data
analysis. Springer New York. Retrieved from
http://had.co.nz/ggplot2/book
Wilkinson, L., Wills, D., Rope, D., Norton, A., & Dubbs, R.
(2006). The grammar of graphics. Springer.