7/13/2020 R for Statistical Analysis

Workshop: R for Statistical Analysis

Data Analysis Team:

Matty Jullamon (GAA)

Amir Michalovich (GAA)
Jeremy Buhler (Data Librarian)
Sarah Parker (Data Librarian)

Pre-workshop setup

Download and install R

For Windows:

1. Visit R Project (https://www.r-project.org/) to learn about R versions.

2. Download and install R from your preferred CRAN mirror here (https://cran.r-project.org/mirrors.html)
A. Choose "0-Cloud" or a mirror site near you.

For Mac:

1. Check that your macOS system is up-to-date

2. Download and install R from The Comprehensive R Archive Network (https://cran.r-project.org/)

Download and install R studio

For Windows and Mac:

1. Download and install R Studio from here (https://rstudio.com/products/rstudio/download/#download)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 1/16

7/13/2020 R for Statistical Analysis

Learning Objectives
Learn how to identify the types of variables in R
Learn the basic commands for descriptive statistics
Learn the basic commands for inference statistics

Overview of Quantative Research Process

A systamatic research process that involves collecting objective, measureable data, using statistics to analyze
the data, and generalizing the results to a larger population to explain a phenomena. Usually, software programs
assist on data analysis.

Data Analysis in Quantitative Research

Definitions

Data refers to facts or pieces of information that can either be quantitative or qualitative.
Variable refers to any property that can be observed or measured.

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 2/16

7/13/2020 R for Statistical Analysis

Types of Variables

It is important to understand the different types of variables because they will determine the statistical analysis
method.

Type Description Example

Nominal 名义上的 Labels or Descriptions that cannot be ordered Gender

Ordinal Labels or Descriptions that can be ordered Education Level

Interval Numeric values with equal magnitude, doesn't have absolute zero SAT scores

Ratio Numeric values with equal magnitude, does have absolute zero Age

Categorize these variables in R

Nominal/Ordinal -> Character or Factor

Intervel/Ratio -> Numeric or Integer

Definitions

Character: Text
Factor: Integer associated with a specific category
Numeric: Number with decimal point
Integer: Number with no decimal point

Getting Started
Set working directory in R studio

You can set the working directory using Session > Set Working Directory > Choose Directory.

Loading a built-in R dataset

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 3/16

7/13/2020 R for Statistical Analysis

About the data

3 Measures Of Ability: SATV, SATQ, ACT: "Self reported scores on the SAT Verbal, SAT Quantitative and ACT
were collected as part of the Synthetic Aperture Personality Assessment (SAPA) web based personality
assessment project. Age, gender, and education are also reported. The data from 700 subjects are included here
as a demonstration set for correlation and analysis" (Revelle et al., 2009).
(https://www.rdocumentation.org/packages/psych/versions/1.9.12.31/topics/sat.act)

Write and run the following commands to load the dataset

install.packages("psych") #install this if you haven't done so.

library(psych)

scores <- sat.act

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 4/16

7/13/2020 R for Statistical Analysis

In [ ]: scores <-read.csv("sat.act.csv")

Identifying and Renaming Variables

str(df ): To check the structure of your data

Question: What do you notice?

as.factor(df$columnname): To change a variable to factor

In [64]: scores$gender <- as.factor(scores$gender)

is.factor(df$columnname): To check if a variable is defined as factor

In [62]: is.factor(scores$gender)

TRUE

Extra information

is.integer(df$columnname): To check if a variable is defined as integer

is.numeric(df$columnname): To check if a variable is defined as numeric

is.character(df$columnname): To check if a varible is defined as character

as.integer(df$columnname): To change a variable to integer

as.numeric(df$columnname): To change a variable to numeric

as.character($columnndfame): To change a variable to character

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 5/16

7/13/2020 R for Statistical Analysis

Exercise #1
Using as.factor command, change 'education' to factor.
Using is.factor command, check if 'education is defined as factor.

Answer to Exercise #1
In [75]: scores$education <- as.factor(scores$education)

In [76]: is.factor(scores$education)

TRUE

Descriptive Statistics

Descriptive statistics summarize the data in a meaningful way. The purpose of using descriptive statistics is to
explore the observed data and not to draw inferences.

We will use the psych package functions to perform descriptive statistics.

describe(df): To obtain descriptive statistics for all variables

hist(df$columnname): To graphically descibe the distribution using histogram

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 6/16

7/13/2020 R for Statistical Analysis

In [78]: hist(scores$SATV)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 7/16

7/13/2020 R for Statistical Analysis

In [95]: # To add the distribution curve

hist(scores$SATV, freq = F)
x <- 200:800
y <- dnorm(x = x, mean = 612.23, sd = 112.90)
lines(x = x, y = y)

Inferential Statistics

Unlike descriptive statistics, inferential statistics use the observed data to make inferences about the population.

In this workshop, we will cover four parametric tests: Independent t-test, One-way ANOVA, Pearson's correlation,
& Simple linear regression. These tests are called parametric because they meet the assumptions of probability
distribution.

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 8/16

7/13/2020 R for Statistical Analysis

Model assumptions
Common model assumptions found in parametric tests:

1. Independence
2. Normality
3. Equal variance

Both Pearson's correlation and Simple linear regression have some additional assumptions. For more
information, click on the following links:

Pearson's correlation assumptions (https://www.statisticssolutions.com/correlation-pearson-kendall-spearman/)

Simple linear regression assumptions (https://www.statisticssolutions.com/assumptions-of-linear-regression/)

Independent T-test
It is used to see whether there are group difference in numeric data between two groups.

For example, do males and females have different average SAT verbal scores?

H0: Mean SATV for males = Mean SATV for females

H1: Mean SATV for males != Mean SATV for females

In [ ]: #Write and run this command:

t.test(scores$SATV ~ scores$gender, data = scores, var.eq = TRUE)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 9/16

7/13/2020 R for Statistical Analysis

Interpreting the results

t: computing statistics
df: degrees of freedom
p-value: Statistical significance. 0.6187 is bigger than α = 0.05 so that means we must retain the null
hypothesis

Conclusion

There was no statistically significant difference in SAT verbal scores between males and females, t(698) = 0.50,
p = 0.62.

One-way ANOVA
It is used to determine whether there are group differences in numeric data between more than two groups

For example, do SAT verbal scores significantly differ by educational levels (1= HS, 2= some college degree, 3 =
2-year college degree, 4= 4-year college degree, 5= graduate work)?

H0: Mean SATV of students who have HS degree = Mean SATV of students who have some college degree = ...

H1: Mean SATV of students who have HS degree != Mean SATV of students who have some college degree !=
...

In [ ]: #Write and run this command:

m1 <- aov(scores$SATV ~ scores$education, data = scores)

summary(m1)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 10/16

7/13/2020 R for Statistical Analysis

Interpreting the results

df: degree of freedom

sum sq: sum of squares
mean sq: mean squares
F value: computing statistics
Pr(>F): statistical significance. 0.275 is bigger than α = 0.05 so that means we must retain the null
hypothesis. We do not have to run the post hoc tests because the group differences are not significant.

Does this make sense?

Check using boxplot.

In [ ]: # Write and run this command:

boxplot(scores$SATV ~ scores$education)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 11/16

7/13/2020 R for Statistical Analysis

Conclusion

There were no significant group differences in SAT verbal scores according to students' educational levels, F(5,
694) = 1.269, p = 0.275.

Extra information

There are different types of post hoc tests

(https://www.rdocumentation.org/packages/DescTools/versions/0.99.36/topics/PostHocTest), but the Tukey's
HSD is the most popular post hoc test for comparing multiple pairings.

In [2]: # R command for Tukey's HSD:

# TukeyHSD(aov(scores$SATV ~ scores$education, data = scores), conf.level=.95)

Pearson's Correlation
It is used to examine relationships between variables (represented by numeric data)

For example, Is there a relationship between SATV and SATQ?

H0: There is no relationship between SATV and SATQ.

H1: There is a relationship between SATV and SATQ.

In [ ]: # Write and run this command:

cor.test(scores$SATV,scores$SATQ)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 12/16

7/13/2020 R for Statistical Analysis

In [ ]: # Write and run this command to see scatterplot:

plot(scores$SATV,scores$SATQ)
abline(lm(scores$SATQ ~ scores$SATV)) #to add regression line

Conclusion

There was statistically significant positive correlation between SAT verbal scores and SAT Quantitative scores (r
= 0.644, p < 0.001).

Simple Linear Regression (for your reference)

It is used to explain/predict the phenomenon of interest based one independent variable.

For example, do ACT scores predict SAT verbal scores?

In [ ]: # Write and run this command:

m2 <- lm(scores$SATV ~ scores$ACT)

summary(m2)

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 13/16

7/13/2020 R for Statistical Analysis

Interpreting the results

The estimated regression line equation: SATV = 237.34 + 13.13(ACT scores)

31% of the variability in the SAT verbal scores was explained by the variables in the regression model.
The overall regression model significantly explained the SAT verbal scores.

Conclusion

ACT scores significantly predicted SAT verbal scores. We would expect 13.13 points increase in SAT verbal
scores for every one point increase in ACT score, assuming all the other variables are held constant.

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 14/16

7/13/2020 R for Statistical Analysis

Questions?

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 15/16

7/13/2020 R for Statistical Analysis

Reference(s);

Revelle, William, Wilt, Joshua, and Rosenthal, Allen. (2009). Personality and Cognition: The Personality-
Cognition Link. In Gruszka, Alexandra and Matthews, Gerald and Szymura, Blazej (Eds.) Handbook of Individual
Differences in Cognition: Attention, Memory and Executive Control, Springer.

https://ubc.syzygy.ca/jupyter/user/mj2653/nbconvert/html/R for Statistical Analysis.ipynb?download=false 16/16