Introduction

We can produce graphics in base R, as demonstrated in the data and descriptive statistics page.

We can also produce superb graphics using the powerful package ggplot2. We focus here on producing the following types of plots:

histograms
scatter plots
box plots, and
bar plots

I encourage you to replicate all the plots in this tutorial in your own RStudio session.

The ggplot2 package is part of the tidyverse package, so begin your session by loading the tidyverse. Recall, to do this, run the line

library(tidyverse)

With the tidyverse loaded, you are ready to use the ggplot commands for generating plots.

All the plots in this tutorial use the earthquakes data set that comes with the openintro package from our text. This data frame contains information about all major 20th century earthquakes.

If you have installed the openintro package, and loaded it into your session, run this line

df <- earthquakes

Alternatively, you can load the data frame into your session directly from its url:

df <- read.csv("https://www.openintro.org/data/csv/earthquakes.csv")

Here are the first three rows of the data frame

head(df,3)

## # A tibble: 3 × 7
##    year month      day richter area                        region    deaths
##   <dbl> <chr>    <dbl>   <dbl> <chr>                       <chr>      <dbl>
## 1  1902 April       19     7.5 Quezaltenango and San Marco Guatemala   2000
## 2  1902 December    16     6.4 Uzbekistan                  Russia      4700
## 3  1903 April       28     7   Malazgirt                   Turkey      3500

The key to using ggplot: A plot begins with the ggplot() command, which is followed by layers describing the plot(s) and features of the plot(s).

Example 1: A histogram of all magnitudes (on the richter scale):

ggplot(data=df)+
  geom_histogram(aes(x=richter),col="white",fill="steelblue",bins=10) +
  ggtitle("Major 20th century Earthquakes") +
  xlab("Magnitutde on the Richter scale")

Notes on code:

The first line says we’re going to make a ggplot using the df data frame. Three layers follow this initial line.
The first layer specifies that we plot a histogram of the variable richter, and adds color to the bars, along with how many bins to make.
The second layer provides a plot title
The third layer provides an x-axis label.

Additional Notes on code:

We specify variables involved in a plot within the aes() command, aes being short for aesthetic. The dots and lines in plots have certain locations, colors, shapes, and sizes. In ggplot, these features are called aesthetics.
Add a + sign at the end of a line if you plan to add another layer.
Plotting data with ggplot requires the data to be within a data frame.

Histograms with ggplot

Key layer: Use geom_histogram(), and specify x inside aes().

Basic Histogram

ggplot(data = df) +
  geom_histogram(aes(x=richter))

Add color and axis labels

ggplot(data = df) +
  geom_histogram(aes(x=richter),bins = 20,col="blue",fill="yellow") + 
  ggtitle("Major 20th century Earthquakes") +
  xlab("Magnitude on the Ricther scale")

The col option colors the boundary of each bar, the fill option colors the interior of each bar. If we want to change the y-axis label, we add the layer ylab("enter new label here inside quotes").

Specify the bins

Specifying the bins for a histogram is good practice. You can either specify the bin widths with the binwidth option inside the geom_histogram() command, or you can specify the total number of bins with the bins option.

Using `binwidth` option

In the following graph, each bin has width 0.4.

ggplot(data = df) +
  geom_histogram(aes(x=richter),col="white",binwidth = 0.4)

Using the `bins` option

In the following plot we create 16 equal width bins

ggplot(data = df) +
  geom_histogram(aes(x=richter),col="white",bins = 16)

Add a vertical line to a histogram

We can add a vertical line layer to a plot with geom_vline(). For instance, we may want to clearly mark in a histogram the mean value of the data.

ggplot(data = df) +
  geom_histogram(aes(x=richter),col="white",fill="dodgerblue",bins=16)+
  geom_vline(aes(xintercept=mean(richter)),
             color="brown", linetype="dashed", size=1)

We can add non-vertical lines to plots as well, and go through this in the scatter plots section of this tutorial.

Change the theme

If you’re not a fan of the gray plot background, you can change the theme. Here are two other options:

ggplot(data=df)+
  geom_histogram(aes(x=richter),col="white",fill="steelblue",bins=10) +
  ggtitle("Major 20th century Earthquakes") +
  xlab("Magnitutde on the Richter scale")+
  theme_classic()

ggplot(data=df)+
  geom_histogram(aes(x=richter),col="white",fill="steelblue",bins=10) +
  ggtitle("Major 20th century Earthquakes") +
  xlab("Magnitutde on the Richter scale") + 
  theme_bw()

Density Plots

Instead of a histogram of counts, we can produce a histogram of relative frequencies by adding the option aes(y = ..density..) as below. This will produce a histogram that records the proportion of the values falling in each bin, not the total counts.

ggplot(df) + 
  geom_histogram(aes(x=richter,y = ..density..), bins=16, col="white", fill="dodgerblue",alpha = 0.4)

Note: The alpha option refers to the opacity of the fill color. Values of alpha range from 0 to 1, with lower values corresponding to more transparent colors.

Sketch a density curve

The geom_density() command gives an idealized density curve rather than a histogram.

ggplot(df) + 
  geom_density(aes(x=richter))

Scatter Plots with ggplot

Key layer: Use the geom_point() plot type command, and specify x and y inside aes().

Basic scatter plot

Although there is likely no association, we can plot earthquake magnitude against the day of the month on which it occurred.

ggplot(data = df) +
  geom_point(aes(x=richter,y=day))

Add color and axis labels

ggplot(data = df) +
  geom_point(aes(x=richter,y=day),col="blue")+
  xlab("magnitude") + 
  ylab("day of the month") +
  ggtitle("Major 20th Century Earthquakes") + 
  theme_bw()

Notes:

In this plot we specify the color (“blue”) outside the aesthetic. This if fine if we want a single color for all the points. If we want the color of a point to depend on some categorical variable, as in the plot below, we specify that inside aes().
R has lots of built-in color names. You can see the names in RStudio if you run colors().

Add color by category

We can color points in a scatter plot according to a categorical variable by specifying col = this variable within the aes() command.

ggplot(data = df) +
  geom_point(aes(x=richter,y=day,col=month))+
  xlab("magnitude") + 
  ylab("day of the month") +
  ggtitle("Major 20th Century Earthquakes")

Bonus Round - Fit the line

We have two ways to add a line to a plot in ggplot.

Using geom_abline()

The first approach is to add the line by specifying the slope and y-intercept using geom_abline(slope = , intercept = ).

For instance, the slope and \(y\)-intercept for the least squares line in the faithful example are 20.227 and -0.4561, respectively (found by using the code lm(day~richter,df)).

Knowing the slope and intercept values, we include the least-squares line in a scatter plot by having two layers in our plot: - a geom_point layer which plots the points, and = a geom_abline layer which plots the line.

ggplot(data = df) +
  geom_point(aes(x=richter,y=day))+
  geom_abline(slope = -0.4561, intercept = 20.227)

Note: We can change the size of the points (size) and the thickness of the line (linewidth), as well as the linetype (to dashed, for instance), by adding these options to their respective layers.

ggplot(data = df) +
  geom_point(aes(x=richter,y=day),size=.5)+
  geom_abline(slope = -0.4561, intercept = 20.227,col="blue",
              linetype = "dashed",
              linewidth = .7)

Using geom_smooth()

The second approach to fitting the least squares regression line to a scatter plot is to use a geom_smooth() layer:

ggplot(data = df,aes(x=richter,y=day)) +
  geom_point(size = .5)+
  geom_smooth(method = 'lm', 
              formula = y~x, 
              se = FALSE) +
  theme_bw()

Note: Now the \(x\) and \(y\) coordinates in the plot are specified within the ggplot command since both the geom_smooth and geom_point commands require them. Alternatively, we could have indicated them in both layers.

Box plots with ggplot

Basic box plot

Key layer command: geom_boxplot()

ggplot(data = df) +
  geom_boxplot(aes(x=richter)) +
  xlab("magnitude") + 
  ggtitle("Major 20th century earthquakes")

This plot is unsatisfying because it gives values on the y-axis, which are meaningless in the context of this box plot. We can hide them:

ggplot(data = df) +
  geom_boxplot(aes(x=richter)) +
  xlab("Magnitude") + 
  ggtitle("Major 20th century earthquakes")+
  theme(axis.ticks.y = element_blank(),
        axis.text.y = element_blank())

Add points to box plots

ggplot(data = df) +
  geom_boxplot(aes(x=richter)) +
  geom_jitter(aes(x=richter,y=0),col="seagreen",height=.1,size=1.2)+
  xlab("Magnitude") +
  ylab("")+
  ggtitle("Major 20th century earthquakes")+
  theme(axis.ticks.y = element_blank(),
        axis.text.y = element_blank())

Note: The geom_jitter() command is the same as the geom_point() with the exception that the computer randomly moves the points a tiny bit (a little “jitter”). This feature is a nice way to see multiple points that might otherwise be stacked right on top of one another. The height=.1 option in the geom_jitter layer means I’m letting the y-coordinate of the point (the height) vary plus or minus .1 units from its actual value.

Side-by-side box plots

ggplot loves to make side-by-side box plots from a data frame that has at least one numeric variable and one categorical variable.

Say we want to compare earthquake magnitudes (which is numeric) by month (which is categorical)!

ggplot(data = df) +
  geom_boxplot(aes(x=richter,y=month)) +
  xlab("magnitude") + 
  ylab("month") + 
  ggtitle("Major 20th century earthquake magnitudes, by month")

Notes on plot code:

In the data frame df, richter is numeric, and month is categorical.
Setting x=richter and y=month inside the geom_boxplot() aestheic groups the data by month and produces a box plot of earthquake magnitudes for each month in month.
We can run the box plots vertically if we switch the roles of \(x\) and \(y\):

ggplot(data = df) +
  geom_boxplot(aes(x=month, y=richter)) +
  xlab("month") + 
  ylab("magnitude") + 
  ggtitle("Major 20th century earthquake magnitudes, by month")

Coloring box plots

We can fill box plots with color according to a categorical variable using the fill option.

ggplot(data = df) +
  geom_boxplot(aes(x=richter,y=month,fill=month)) +
  xlab("magnitude") + 
  ylab("month") + 
  ggtitle("Major 20th century earthquakes")

Ordering months chronologically

The first line of code below redefines the month column in a way that specifies the months in the correct order.

We can also hide the legend if it is superfluous, as it is in this case, by adding show.legend = FALSE inside the geom_boxplot() command.

df$month = factor(df$month, levels=month.name)
ggplot(data = df) +
  geom_boxplot(aes(x=richter,y=month,fill=month),show.legend=FALSE) +
  xlab("magnitude") + 
  ylab("month") + 
  ggtitle("Major 20th century earthquakes")

Bonus Round - Color Opacity

As mentioned in the histogram section, the alpha option adjusts the opacity of a color in a plot. The closer alpha is to 0, the more transparent it becomes, and the closer to 1, the more opaque it becomes. In the graph below the three box plots are all filled with the color “orange” but with different alpha values.

ggplot(data = df) +
  geom_boxplot(aes(x=richter,y=month),fill="orange",alpha=seq(from=0,to=1,by=1/11),show.legend=FALSE) +
  xlab("magnitude") + 
  ylab("month") + 
  ggtitle("Major 20th century earthquakes")

Bar plot for a categorical variable

In Base R we can visualize the frequencies for a categorical variable as follows:

barplot(table(df$month))

We can also create a bar plot with gglplot with the following code:

ggplot(data=df,aes(x=month))+
  geom_bar(stat="count")

Note: We can specify colors manually, change bar width, add labels, and even rotate them so they look less crowded. We can also specify the limits of the values on the y-axis to be, say, 0 to 20:

colors=c("wheat","steelblue","tomato","seagreen","yellow","purple",
         "pink","orange","brown3","dodgerblue","gray","darkblue")
ggplot(data=df,aes(x=month))+
  geom_bar(stat="count", width=.5, fill=colors)+
  ylim(0,20)+
  ggtitle("Number of earthquakes by month")+
  theme(axis.text.x=element_text(angle=60,vjust=.5))

Cheat Sheet

The following page has a downloadable ggplot cheat sheet (pdf)

MATH 140: Data Visualization with ggplot

Introduction

Histograms with ggplot

Basic Histogram

Add color and axis labels

Specify the bins

Using binwidth option

Using the bins option

Add a vertical line to a histogram

Change the theme

Density Plots

Sketch a density curve

Scatter Plots with ggplot

Basic scatter plot

Add color and axis labels

Add color by category

Bonus Round - Fit the line

Using geom_abline()

Using geom_smooth()

Box plots with ggplot

Basic box plot

Add points to box plots

Side-by-side box plots

Coloring box plots

Ordering months chronologically

Bonus Round - Color Opacity

Bar plot for a categorical variable

Cheat Sheet

Using `binwidth` option

Using the `bins` option