Creating and styling boxplots

Boxplots (also called box-and-whisker plots) are a graphical tool used to summarise and display the distribution of a continuous variable. They are useful for several reasons:

Identifying Outliers: Boxplots clearly highlight outliers (values that fall significantly outside the range of most of the data). Outliers are shown as individual points beyond the “whiskers” of the plot.
Visualizing the Spread and Central Tendency: The box itself shows the interquartile range (IQR). The line inside the box represents the median, providing a sense of central tendency.
Displaying the Range of Data: The boxplot gives a quick overview of the minimum, first quartile (Q1), median (Q2), third quartile (Q3), and maximum values, helping to understand the range and overall distribution.

Key Components of a Boxplot:

Box: Represents the interquartile range (IQR).
Whiskers: Extend to the smallest and largest data points within 1.5 times the IQR from Q1 and Q3.
Median: The line inside the box, representing the middle value of the dataset.
Outliers: Shown as individual points beyond the whiskers.

Boxplots provide a concise summary, making it easier to understand the distributional properties of a dataset at a glance.

In this tutorial, we demonstrate how to create boxplots using the boxplot() function in base R. We will use a dataset from a nutrition survey of school children 10 years and older from Pakistan. This dataset is available from the Oxford iHealth teaching datasets repository.

## link to CSV from GitHub repository ----
1csv_file_url <- "https://raw.githubusercontent.com/OxfordIHTM/teaching_datasets/refs/heads/main/school_nutrition.csv"

## Read CSV file ----
2nut_data <- read.csv(file = csv_file_url)

1: This URL can be retrieved from GitHub by accessing the raw version of the GitHub link to the file
2: Use read.csv() to read the CSV file from the URL

On inspection of the dataset, we see:

## Show first 5 rows of data ----
head(nut_data)

  region school age_months sex weight height
1      1      1        121   2   20.6  124.6
2      1      1        121   1   27.9  130.7
3      1      1        129   2   25.7  131.4
4      1      1        133   1   27.0  135.7
5      1      1        145   2   28.5  130.5
6      1      1        148   2   35.1  142.1

We have a data.frame with 267 rows and 6 columns.

For this tutorial, we will focus on the weight variable in the dataset for demonstrating how to create and style boxplots in base R.

Creating a boxplot

## Boxplot of weight of all children ----
boxplot(nut_data$weight)

Figure 1: Boxplot of weight of all children

By default, the boxplot() function provides the plot above as output when provided values of a continuous variable. Note the default settings in terms of fill for the box and no plot labels.

We might be interested in comparing the distribution of the weight variable between male and female in the dataset. This can be done in different ways using the boxplot() function.

Default method

The default method for creating boxplots for different groupings in a dataset is to supply the different vectors of values for each group either as unnamed arguments or as a single list.

The following code shows one way to provide the different values of weight based on sex as unnamed arguments.

## Default boxplot method using unnamed arguments ----
boxplot(
1  nut_data$weight[nut_data$sex == 1],
2  nut_data$weight[nut_data$sex == 2]
)

1: Indexing the weight variable to get values for males
2: Indexing the weight variable to get values for females

Figure 2: Boxplot of weight by sex - default method using unnamed arguments

The following code shows one way to provide a single list of values of weight for males and then females.

## Default boxplot method using single list ----
1boxplot(split(nut_data$weight, nut_data$sex))

1: Use of split() function to split the weight values by sex. The split() function outputs a single list with weight values grouped by sex.

Figure 3: Boxplot of weight by sex - default method using single list

Both approaches produce the same plot. Each boxplot is labeled according to the values of the grouping variable which in this case is sex with 1 corresponding to males and 2 corresponding to females.

Formula method

The boxplot() function has a formula method for plotting grouping for a variable of interest as shown below:

## Boxplot of weight by sex using formula method ----
boxplot(
1    weight ~ sex,
2    data = nut_data
)

1: This is the standard formula specification for the weight variable being grouped by the sex variable. We use the ~ operator to express the formula.
2: For formula method, the data argument is required and this is where you provide the name of the object containing the dataset that has the variable weight and sex used in the formula

Figure 4: Boxplot of weight by sex - formula method

With the formula method, the output is similar with the previous approach with the only difference being the x and y labels have been set using the values in the formula. From a syntax perspective, the formula method is more compact without requiring any other processing of the input dataset.

Styling a boxplot

The boxplots we’ve created so far are quite basic and would benefit from a little bit more styling. At the minimum, we would like to:

Add a title to the plot;
Add more informative labels to each boxplot;
Add x- and y-axis labels (or update them to look more presentable and complete); and,
Modify some plot layout features such as box width and plot frames as appropriate.

Adding a title

Since the boxplot() functions is built on the generic plot() base function in R, named graphical parameters/arguments used in plot() can also be specified in boxplot(). So, to add a title to a boxplot, we do this:

boxplot(
  nut_data$weight,
1  main = "Summary of weight values for all children"
)

1: We use the graphical parameter main to set the title of the boxplot

To add a title to the the boxplot for weight by sex, we do this:

boxplot(
  weight ~ sex,
  data = nut_data,
  main = "Summary of weight values by sex of children"
)

Figure 6: Boxplot of weight by sex with title

Adding names to each boxplot

For the boxplot of weight by sex, it would be more appropriate for the labels under each boxplot to be more informative of the values they represent. Here, instead of 1 and 2, it would be better to have “Male” and “Female” as labels. We can do this as follows:

boxplot(
  weight ~ sex,
  data = nut_data,
  main = "Summary of weight values by sex of children",
1  names = c("Male", "Female")
)

1: The names argument in the boxplot() function is used to give different labels/values to the different boxplots instead of the values provided by the data.

Figure 7: Boxplot of weight by sex with title and boxplot labels

Adding x and y axis labels

We can again use generic graphical parameters to set x and y axis labels to a boxplot as follows:

boxplot(
  nut_data$weight,
  main = "Summary of weight values for all children",
1  ylab = "Weight (kgs)"
)

1: We use the ylab graphical parameter to specify a label for the y-axis.

Figure 8: Boxplot of weight with title and x and y labels

For the boxplot for weight values for all children, an x axis label is not really necessary.

We can set the x and y axis labels to the boxplot for weight by sex as follows:

boxplot(
  weight ~ sex,
  data = nut_data,
  main = "Summary of weight values by sex of children",
1  xlab = "Sex",
  ylab = "Weight (kgs)",
  names = c("Male", "Female")
)

1: We use the xlab graphical parameter to specify a label for the x-axis.

Figure 9: Boxplot of weight by sex with title and x and y labels

Modifying other plot layout options

Further modifications can be applied to the boxplot that may enhance their style. The following code adjusts the width of the boxplots and removes the plot frame.

boxplot(
  weight ~ sex,
  data = nut_data,
  main = "Summary of weight values by sex of children",
  xlab = "Sex",
  ylab = "Weight (kgs)",
  names = c("Male", "Female"),
1  boxwex = 0.5,
2  frame.plot = FALSE
)

1: Set value for boxwex parameter to scale the size of the boxes. This is particularly useful when there are only a few groups to create boxplots of.
2: Set frame.plot graphical parameter to FALSE to remove the plot frame.

Figure 10: Boxplot of weight by sex with title and labels and with modified layout