Home

/

GitHub

/

In jr-packages/jrGgplot2: Jumping Rivers: Advanced Graphics in R

If you attended the Introduction to R course with us, you will already be familiar with some of the basic ggplot2 concepts. This practical serves as a reminder on some of those concepts, whilst also introducing some new ones. If you didn't attend this course, use this practical as a introduction to the basic concepts of ggplot2. Some these of plots aren't particularly useful, we are just using them for illustration purposes.

\newthought{To begin with}, load the ggplot2 package^[The ggplot2 package is automatically installed with jrGgplot2.]

library("ggplot2")

\noindent Next we load the beauty data set:^[Details of the beauty data set can be found at the end of this practical.]

library("jrGgplot2")
data(Beauty, package = "jrGgplot2")

\noindent When loading in data, it's always a good idea to carry out a sanity check. I tend to use the commands

head(Beauty)
colnames(Beauty)
dim(Beauty)

Scatter plots

Scatter plots are created using the point geom. Let's start with a basic scatter plot

ggplot(data = Beauty) +
  geom_point(aes(x = age, y = beauty))

\noindent To save typing, we can also store the plot as a variable:^[In this practical, we are creating the plots in a slightly verbose way.]

g = ggplot(data = Beauty)
g1 = g + geom_point(aes(x = age, y = beauty))

\noindent To view this plot, type g1.

The arguments x and y are called aesthetics. For geom_point, these parameters are required. This particular geom has other aesthetics: shape, colour, size and alpha.^[These aesthetics are usually available for most geoms.] Here are some things to try out.

Experiment with other aesthetics. For example,

g + geom_point(aes(x = age, y = beauty, colour = gender))

\noindent or

g + geom_point(aes(x = age, y = beauty, colour = gender,
                   alpha = evaluation))

Some aesthetics, like shape must be discrete. So we have to transform the variable into a character or factor - shape = factor(tenured). - Are there any differences between numeric values like tenured and characters like gender for some aesthetics? What happens if you convert tenured to a factor in the colour aesthetic. For example, colour = factor(tenured). - What happens if you set colour (or some other aesthetic) outside of the aes function? For example, compare

g + geom_point(aes(x = age, y = beauty, colour = "blue"))

\noindent to

g + geom_point(aes(x = age, y = beauty), colour = "blue")

What happens when you set an aesthetic to a constant value. For example, colour = 2. What happens if you put this argument outside of the aes function?

Box plots

The box plot geom has the following aesthetics: x, y, colour, fill, linetype, weight, size and alpha. We can create a basic boxplot using the following commands:

g + geom_boxplot(aes(x = gender, y = beauty))

\noindent Similar to the point geom, we can add in aesthetics:

g + geom_boxplot(aes(x = gender, y = beauty,
                     colour = factor(tenured)))

\noindent Why do you think we have to convert tenured to a discrete factor?

As before, experiment with the different aesthetics. For some of the aesthetics, you will need to convert the continuous variables to discrete variables. For example, this will give an error:

g + geom_boxplot(aes(x = gender, y = beauty, colour = tenured))

\noindent while this is OK^[Why?]

g + geom_boxplot(aes(x = gender, y = beauty,
                     colour = factor(tenured)))

\noindent Make sure you play about with the different aesthetics.

Combining plots

The key idea with ggplot2 is to think in terms of layers not in terms of plot "types".^[In the lectures we will discuss what this means.] For example,

g + geom_boxplot(aes(x = gender, y = beauty,
                          colour = factor(tenured))) +
  geom_point(aes(x = gender, y = beauty))

What happens to the plot if you swap the order of the geom_boxplot and geom_point function calls?
In this case, geom_point isn't that great. Try using geom_jitter^[We have a bit too much data for geom_jitter, but you get the point.]:

g + geom_boxplot(aes(x = gender, y = beauty,
                         colour = factor(tenured))) +
  geom_jitter(aes(x = gender, y = beauty))

Bar plots

The bar geom has the following aesthetics: x, colour, fill, size, linetype, weight and alpha. Here is a command to get started:

g + geom_bar(aes(x = factor(tenured)))

As before, try different aesthetic combinations. Convert parameters to discrete versions as needed using factor(...).
Let's get a bit more fancy. First, we round ages to the nearest decade:

Beauty$dec = factor(signif(Beauty$age, 1))

\noindent then plot:

g = ggplot(data = Beauty)
g + geom_bar(aes(x = gender, fill = dec))

\noindent We can adjust the layout of this bar plot using ggplot's position adjustments. The five possible adjustments are listed in table 1. The default adjustment is stack

g + geom_bar(aes(x = gender, fill = dec),
             position = "stack")

Try the other adjustments.

g + geom_bar(aes(x = gender, fill = dec),
             position = "dodge")

\begin{table}[t] \centering \begin{tabular}{@{}ll@{}} \toprule Adjustment & Description \ \midrule \texttt{dodge} & Adjust position by overlapping to the side \ \texttt{fill} & Stack overlapping elements; standardise stack height\ \texttt{identity} & Do nothing \ \texttt{jitter} & Jitter points \ \texttt{stack} & Stack overlapping elements \ \bottomrule \end{tabular} \caption{Position adjustments - table 4.5 in the ggplot2 book.} \label{T1} \end{table}

\newpage

The beauty data set

\begin{table}[!t] \centering \caption{The first five rows of the beauty data set. There are a total of 463 course evaluations.} \begin{tabular}{@{}llllll r@{.}l@{}} \toprule tenured & minority & age & evaluation & gender & students & \multicolumn{2}{l}{beauty} \ \midrule 0 & 1 & 36 & 4.3 & Female & 43 & 0&202 \ 1 & 0 & 59 & 4.5 & Male & 20 & -0&826 \ 1 & 0 & 51 & 3.7 & Male & 55 & -0&660 \ 1 & 0 & 40 & 4.3 & Female & 46 & -0&766 \ 0 & 0 & 31 & 4.4 & Female & 48 & 1&421 \ \bottomrule \end{tabular} \label{T2} \end{table}

This data set is from a study where researchers were interested in whether a lecturers' attractiveness affected their course evaluation.\cite{Hamermesh2003} This is a cleaned version of the data set and contains the following variables:

evaluation - the questionnaire result.
tenured - does the lecturer have tenure; 1 == Yes. In R, this value is continuous.
minority - does the lecturer come from an ethnic minority (in the USA).
age.
gender.
students - number of students in the class.
beauty - each of the lecturers’ pictures was rated by six undergraduate students: three women and three men. The raters were told to use a 10 (highest) to 1 rating scale, to concentrate on the physiognomy of the professor in the picture, to make their ratings independent of age, and to keep 5 in mind as an average. The scores were then normalised.