Data Visualization Analysis with R

(v. 2.0)

Oscar Torres-Reyna
[email protected]

June 2019 http://dss.princeton.edu/training/

 Data visualization helps reduce the mental
stress of extracting meaning from data and
plays an important role at all stages of data
analysis when exploring data, making
inferences, and presenting results

DSS/OTR 2
 While different fields of study have
developed their own way of visualizing data,
the common goal across all types of visuals
for data analysis is to find meaningful
patterns through trends, relationships, or
distribution.
See the following site
https://www.r-graph-gallery.com/

DSS/OTR 3
The data on the right shows a snapshot
of the unemployment rate and
presidential approval in the United
States from 1948 to 2019.
In table form it is hard to figure
something out of the data, except for
the fact that it represents indicators
measured over time, making it an ideal
candidate for a time series graph.

DSS/OTR 4
 We will be using RStudio, please see the
following document for some introduction to
its interface

https://dss.princeton.edu/training/RStudio101.pdf

DSS/OTR 5
 We shall start by activating some of the R
packages we will need for this document:
library(zoo)
library(ggplot2)
library(stargazer)

If the package is not available, you need to

install it, type
install.packages("name of package")

DSS/OTR 6
The data is in *.csv format, we can use the read.csv() function
import it into R:

mydata = read.csv("http://www.princeton.edu/~otorres/mydataviz.csv",
header = TRUE, stringsAsFactors = FALSE)

1) unemp = Unemployment from the Bureau of Labor Statistics:

2) approve = Own estimation of monthly averages using presidential

approval data from ROPER center. Note that data is publicly
available at The American Presidency Project at the following site:
https://www.presidency.ucsb.edu/statistics/data/presidential-job-approval

DSS/OTR 7
mydata$date = as.Date(mydata$date,"%m/%d/%Y")

mydata = mydata[ order(mydata$date),]

mydata$month = as.yearmon(mydata$date)

DSS/OTR See: https://dss.princeton.edu/training/TimeSeriesR101.pdf 8

# Plotting unemployment data using base R function plot()
plot(mydata$month, mydata$unemp,
type = "l", main="Unemployment", ylim=c(0,12),
xlab="Month", ylab="Unemployment rate (%)")

DSS/OTR 9
# Plotting unemployment data using ggplot2
ggplot(data=mydata, aes(x=month, y=unemp)) +
geom_line() +
labs(title = "US Unemployment rate, 1948-2019 (monthly)",
y = "Unemployment rate (%)",
x = "Monthly",
caption = "Source: Unemployment data from the BLS") +
scale_y_continuous(limits = c(0, 12), breaks = seq(0,12))

DSS/OTR 10
We can add analytic component to the visual by incorporating context to the
trends. In the next slides we will see the unemployment trends by presidential
terms.

# Mid-point date per administration

terms = data.frame(month = as.yearmon(c("Jan 1950","Jan 1957","Jan 1962","Jan 1966","Jan 1971",

"Sep 1975","Jan 1979","Jan 1985","Jan 1991","Jan 1997",
"Jan 2005","Jan 2013","Jan 2019")),
y = 12,
name = c("Truman(D)","Eisenhower(R)","Kennedy(D)","L. Johnson(D)","Nixon(R)",
"Ford(R)","Carter(D)","Reagan(R)","G. Bush(R)","Clinton(D)",
"G.W. Bush(R)","Obama(D)","Trump(R)"))

# Begin date

start = c("Jan 1948","Jan 1953","Jan 1961","Jan 1963",

"Jan 1969","Aug 1974","Jan 1977","Jan 1981",
"Jan 1989","Jan 1993","Jan 2001","Jan 2009","Jan 2017")

# End date

end = c("Jan 1953","Jan 1961","Jan 1963","Jan 1969",

"Aug 1974","Jan 1977","Jan 1981","Jan 1989",
"Jan 1993","Jan 2001","Jan 2009","Jan 2017","Dec 2019")

# Order

order = c("blue","red","blue","blue","red","red","blue","red","red","blue","red","blue","red")

DSS/OTR 11
mydata$unempd = ifelse(mydata$party=="Democrat",mydata$unemp, NA)
mydata$unempr = ifelse(mydata$party=="Republican",mydata$unemp, NA)

ggplot(data=mydata, aes(x=month)) +
geom_line(data = mydata, aes(y = unempd), color = "blue", size = 1) +
geom_line(data = mydata, aes(y = unempr), color = "red", size = 1) +
See next page
theme_bw() +
labs(title = "US Unemployment rate, 1948-2019 (monthly)",
y = "Unemployment rate (%)",
x = "Monthly",
caption = "Source: Unemployment data from the BLS") +
theme(plot.title = element_text(hjust = 0.5)) +
scale_y_continuous(limits = c(0, 12), breaks = seq(0,12)) +
scale_x_yearmon(breaks = as.yearmon(start)) +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
geom_vline(xintercept = as.yearmon(start),
linetype = "solid",
size = 0.5,
color = "black") +
annotate("rect",
xmin = as.yearmon(start),
xmax = as.yearmon(end),
ymin = -Inf, ymax = Inf,
fill = order,
alpha = 0.2) +
geom_text(data = terms, aes(x=month, y = y, label=name), angle = 270, hjust = 0, fontface= "bold")

DSS/OTR 12
DSS/OTR 13
 Adding presidential
approval
See next page
ggplot(data=mydata, aes(x=month)) +
geom_line(data = mydata, aes(y = unempd), color = "blue", size = 1) +
geom_line(data = mydata, aes(y = unempr), color = "red", size = 1) +
theme_bw() +
labs(title = "US Unemployment rate and Presidential Approval, 1948-2019 (monthly)",
y = "Unemployment rate (%)",
x = "Monthly",
caption = "Source: Unemployment data from the BLS, Presidential Approval from ROPER") +
theme(plot.title = element_text(hjust = 0.5)) +
scale_y_continuous(limits = c(0, 12), breaks = seq(0,12)) +
scale_x_yearmon(breaks = as.yearmon(start)) +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
geom_vline(xintercept = as.yearmon(start),
linetype = "solid",
size = 0.5,
color = "black") +
annotate("rect",
xmin = as.yearmon(start),
xmax = as.yearmon(end),
ymin = -Inf, ymax = Inf,
fill = order,
alpha = 0.2) +
geom_hline(yintercept = 50/7.997222) +
geom_text(data = terms, aes(x=month, y = y, label=name), angle = 270, hjust = 0, fontface= "bold") +
geom_line(data = mydata, aes(y = approve/7.997222)) +
scale_y_continuous(sec.axis = sec_axis(~.*7.997222,(name = "Presidential Approval (%)")),
limits = c(0, 12), breaks = seq(0,12))

DSS/OTR 14
DSS/OTR 15
The previous visual does not provide a clear idea of the relationship between unemployment
rates and presidential approval. Scatterplots are ideal plots to find relationships between
variables. The following code produces a scatterplot using base R:
plot(mydata$unemp, mydata$approve,
main = "US Unemployment rate and Presidential Approval, 1948-2019 (monthly)",
xlab = "Unemployment rate (%)", ylab = "Presidential Approval (%)")

DSS/OTR 16
The car package provides an informative scatterplot including a linear and loess fit with
boxplots
scatterplot(approve ~ unemp, data = mydata,
main = "US Unemployment rate and Presidential Approval, 1948-2019 (monthly)",
xlab = "Unemployment rate (%)", ylab = "Presidential Approval (%)")

DSS/OTR 17
Per group
scatterplot(approve ~ unemp|party, data = mydata,
main = "US Unemployment rate and Presidential Approval, 1948-2019 (monthly)",
xlab = "Unemployment rate (%)", ylab = "Presidential Approval (%)",
col = c("blue","red"),
legend = c(title="Party", coords="topright"))

DSS/OTR 18
Scatterplots using ggplot()
ggplot(data = mydata, aes(x=unemp, y=approve, group=factor(party), color = factor(party))) +
geom_point(size = 3) +
scale_color_manual(name="",values=c('blue', 'red')) +
theme(legend.position = "bottom") +
labs(title = "US Unemployment rate and Presidential Approval, 1948-2019 (monthly)",
x = "Unemployment rate (%)",
y = "Presidential Approval (%)",
caption = "Source: Unemployment data from the BLS, Presidential Approval from ROPER")

DSS/OTR 19
Adding a linear fit per party
ggplot(data = mydata, aes(x=unemp, y=approve, group=factor(party), color = factor(party))) +
geom_point(size = 3) +
scale_color_manual(name="",values=c('blue', 'red')) +
theme(legend.position = "bottom") +
labs(title = "US Unemployment rate and Presidential Approval, 1948-2019 (monthly)",
x = "Unemployment rate (%)",
y = "Presidential Approval (%)",
caption = "Source: Unemployment data from the BLS, Presidential Approval from ROPER") +
stat_smooth(method=lm)

DSS/OTR 20
We can estimate the equation of the linear fit per party by using the lm()
function and produce a nice presentation using stargazer().

reg1 = lm(approve ~ unemp, data = subset(mydata, party=="Democrat"))

reg2 = lm(approve ~ unemp, data = subset(mydata, party=="Republican"))

stargazer(reg1, reg2, type = "html", out="reg.html",

dep.var.labels=c("Presidential Approval"),
model.numbers = FALSE,
column.labels =c("Democrats", "Republicans"),
covariate.labels=c("Unemployment rate"))

The file reg.html can be opened with Word, it is saved in the working directory.

See: https://dss.princeton.edu/training/Regression101R.pdf and

https://dss.princeton.edu/training/NiceOutputR.pdf
See next page
DSS/OTR 21
DSS/OTR 22