A donut chart (a.k.a. ring chart) is nothing more than a pie chart with a hole inside. This seems pretty obvious. and the code in ggplot2 that draws a donut chart reflects well the simplicity of the problem. If you remember that a pie chart is just a stacked bar plot in polar coordinate system (as described here), then adding a hole to a pie chart requires only that we add a bit of space to the left side of the stacked bar before applying the polar coordinate system!

Let’s see that with the example previously used here for drawing the pie chart. The corresponding variables and dataframe are described here:

# variable 1
type <- c("O+", "A+", "B+", "AB+", "O-", "A-", "B-", "AB-")
# variable 2
proportion <- c(33.2,   41.6,   6.8,    3.4,    5.8,    7.4,    1.2,    0.6)
# dataframe
df <- data.frame(type, proportion)



In this example, we will decompose the procedure into four steps:



1. Building the stacked bar chart

We use geom_col() to draw the bar plot and we map the variable with aes(). Here, x= is set but not named by the mean of "", while y = is set to proportion. We use fill= to attribute a specific color to each blood type, and apply the color palette viridis to use a better set of colors than those provided automatically by ggplot():

ggplot(df, aes(x="", y=proportion, fill=type))+
  geom_col() +
  scale_fill_viridis_d()



2. Adding some space to the left side of the bar.

Now we need to add some space next to the stacked bar. First we set up both position and width for the bar. For example, we set the position of the bar to 2 on the X-axis by writing aes(x=2). Then we set the bar width to 1 with width=1, meaning that the limits of the bar relative to the X-axis are 1.5 and 2.5. If we now restrict the display of the X-axis to the range 0.5-2.5 with xlim(0.5,2.5), we end up with an empty space and the stacked bar both of a width of 1 next to each other:

ggplot(df, aes(x=2, y=proportion, fill=type))+
  geom_col(width=1) +
  xlim(0.5, 2.5) +
  scale_fill_viridis_d()



3. Applying the coordinate system

Now we need to exchange the cartesian coordinate system in a polar coordinate system. In this way, the stacked bar will be stretched around a central point of focus. For that we add coord_polar("y") to the code:

ggplot(df, aes(x=2, y=proportion, fill=type))+
  geom_col(width=1) +
  xlim(0.5, 2.5) +
  scale_fill_viridis_d() +
  coord_polar("y")

As you see here, the empty space to the left of the bar becomes the empty space at the center of the pie chart, thus creating the donut chart.

4. Getting rid of the useless axis

The circular axis around the chart has very little meaning; it may thus be taken away. One way to do it is to remove the whole theme, which means that the grey background disappears as well. To do so, we just add theme_void:

ggplot(df, aes(x=2, y=proportion, fill=type))+
  geom_col(width=1) +
  xlim(0.5, 2.5) +
  scale_fill_viridis_d() +
  coord_polar("y") +
  theme_void()