Box plot by group in ggplot2

Sample data

In this tutorial we are going to use the following data frame, which first column is as numerical variable and the second a categorical variable representing groups.

# Sample dataset
set.seed(136)
df <- data.frame(y = rnorm(300),
                 group = sample(LETTERS[1:3],
                                size = 300,
                                replace = TRUE))

Box plot by group with `geom_boxplot`

In order to create a basic grouped box plot in R you need to pass the variables to aes and use the geom_boxplot geom as in the following example.

# install.packages("ggplot2")
library(ggplot2)

# Box plot by group
ggplot(df, aes(x = group, y = y)) + 
  geom_boxplot()

Box plot by group in ggplot2

Adding error bars with `stat_boxplot`

Box plot by group with error bars in ggplot2

As you may have noticed, the default box plot in ggplot2 doesn’t add error bars. If you want to add them use the stat_boxplot stat and set geom = "errorbar". The width of the bars can be customized with width argument.

# install.packages("ggplot2")
library(ggplot2)

# Box plot by group with error bars
ggplot(df, aes(x = group, y = y)) + 
  stat_boxplot(geom = "errorbar", # Error bars
               width = 0.25) +    # Bars width
  geom_boxplot()

Horizontal box plot by group

The box plots can also be displayed in horizontal or landscape mode. To accomplish it you can change the order of your variables inside aes or use coord_flip, as shown above.

Option 1: change the order of the variables

Set the grouping variable on the Y-axis.

# install.packages("ggplot2")
library(ggplot2)

# Horizontal box plot in ggplot2
ggplot(df, aes(x = y, y = group)) + 
  stat_boxplot(geom = "errorbar",
               width = 0.25) + 
  geom_boxplot()

Horizontal grouped box plot in ggplot

Option 2: use coord_flip

coord_flip will invert the axes, so a vertical box plot will become horizontal and viceversa.

# install.packages("ggplot2")
library(ggplot2)

# Horizontal box plot
ggplot(df, aes(x = group, y = y)) + 
  stat_boxplot(geom = "errorbar",
               width = 0.25) + 
  geom_boxplot() +
  coord_flip()

Horizontal box plot by group in ggplot2

Color customization

Box plot by group with custom color palette in ggplot

If you pass the categorical variable to the fill argument of aes, each box plot will be filled with a color and a legend will be displayed.

# install.packages("ggplot2")
library(ggplot2)

ggplot(df, aes(x = group, y = y, fill = group)) + 
  stat_boxplot(geom = "errorbar",
               width = 0.25) + 
  geom_boxplot()

The colors or the box plots are fully customizable. In the following example we are setting a fill color for each group, changing the border color of the boxes and setting the color of the outliers to black.

Box plot by group with custom color palette in ggplot

# install.packages("ggplot2")
library(ggplot2)

# Fill colors
cols <- c("#CFD8DC", "#90A4AE", "#455A64")

ggplot(df, aes(x = group, y = y, fill = group)) + 
  stat_boxplot(geom = "errorbar",
               width = 0.25) + 
  geom_boxplot(alpha = 0.8,          # Fill transparency
               colour = "#474747",   # Border color
               outlier.colour = 1) + # Outlier color
  scale_fill_manual(values = cols)   # Fill colors

Legend customization

Change the title

You can change the default title of the legend with guides function as follows.

# install.packages("ggplot2")
library(ggplot2)

ggplot(df, aes(x = group, y = y, fill = group)) + 
  stat_boxplot(geom = "errorbar", width = 0.25) + 
  geom_boxplot() +
  guides(fill = guide_legend(title = "Title"))

Change the legend of the box plot in ggplot2

Change the labels

The legend labels are the levels of the categorical variable. You can override them with scale_fill_hue.

# install.packages("ggplot2")
library(ggplot2)

ggplot(df, aes(x = group, y = y, fill = group)) + 
  stat_boxplot(geom = "errorbar", width = 0.25) + 
  geom_boxplot() +
  scale_fill_hue(labels = c("G1", "G2", "G3"))

Change the labels of the legend of the box plot in ggplot2

Remove the legend

If you don’t want any legend displayed you can remove it with theme(legend.position = "none").

# install.packages("ggplot2")
library(ggplot2)

ggplot(df, aes(x = group, y = y, fill = group)) + 
  stat_boxplot(geom = "errorbar", width = 0.25) + 
  geom_boxplot() +
  theme(legend.position = "none")

Remove the box plot legend in ggplot2