R: Pie chart with percentage as labels using ggplot2

Question

From a data frame I want to plot a pie chart for five categories with their percentages as labels in the same graph in order from highest to lowest, going clockwise.

My code is:

League<-c("A","B","A","C","D","E","A","E","D","A","D")
data<-data.frame(League) # I have more variables 

p<-ggplot(data,aes(x="",fill=League))
p<-p+geom_bar(width=1)
p<-p+coord_polar(theta="y")
p<-p+geom_text(data,aes(y=cumsum(sort(table(data)))-0.5*sort(table(data)),label=paste(as.character(round(sort(table(data))/sum(table(data)),2)),rep("%",5),sep="")))
p

I use

cumsum(sort(table(data)))-0.5*sort(table(data))

to place the label in the corresponding portion and

label=paste(as.character(round(sort(table(data))/sum(table(data)),2)),rep("%",5),sep="")

for the labels which is the percentages.

I get the following output:

Error: ggplot2 doesn't know how to deal with data of class uneval

Answer 1

I've preserved most of your code. I found this pretty easy to debug by leaving out the coord_polar... easier to see what's going on as a bar graph.

The main thing was to reorder the factor from highest to lowest to get the plotting order correct, then just playing with the label positions to get them right. I also simplified your code for the labels (you don't need the as.character or the rep, and paste0 is a shortcut for sep = "".)

League<-c("A","B","A","C","D","E","A","E","D","A","D")
data<-data.frame(League) # I have more variables 

data$League <- reorder(data$League, X = data$League, FUN = function(x) -length(x))

at <- nrow(data) - as.numeric(cumsum(sort(table(data)))-0.5*sort(table(data)))

label=paste0(round(sort(table(data))/sum(table(data)),2) * 100,"%")

p <- ggplot(data,aes(x="", fill = League,fill=League)) +
  geom_bar(width = 1) +
  coord_polar(theta="y") +
  annotate(geom = "text", y = at, x = 1, label = label)
p

The at calculation is finding the centers of the wedges. (It's easier to think of them as the centers of bars in a stacked bar plot, just run the above plot without the coord_polar line to see.) The at calculation can be broken out as follows:

table(data) is the number of rows in each group, and sort(table(data)) puts them in the order they'll be plotted. Taking the cumsum() of that gives us the edges of each bar when stacked on top of each other, and multiplying by 0.5 gives us the half the heights of each bar in the stack (or half the widths of the wedges of the pie).

as.numeric() simply ensures we have a numeric vector rather than an object of class table.

Subtracting the half-widths from the cumulative heights gives the centers each bar when stacked up. But ggplot will stack the bars with the biggest on the bottom, whereas all our sort()ing puts the smallest first, so we need to do nrow - everything because what we've actually calculate are the label positions relative to the top of the bar, not the bottom. (And, with the original disaggregated data, nrow() is the total number of rows hence the total height of the bar.)