Aggregating data with ggplot

Question

Let's use the mpg dataset as an example, specifically the class and cyl columns. I can see how many entries are there, per single class, and differentiate the fill color based on the cyl value:

library(ggplot2)
p <- ggplot(mpg)
p <- p + geom_bar(mapping=aes(x=class, fill=factor(cyl)), position=position_dodge())
print(p)

What I'd like to see, though, is the average count of entries (per class), per different values of cyl. Basically, if you look at the plot above, I want a single bar per class, whose height should be the average height of the colored bars for that class.

I am able to get this result by preprocessing the data frame, e.g.:

df <- aggregate(formula=cyl~class, data=mpg, FUN=function(x) { length(x) / length(unique(x)) })
p <- ggplot(df)
p <- p + geom_bar(mapping=aes(x=class, y=cyl), stat='identity')
p <- p + ylab('average count')

That gives my desired output:

However, given how powerful ggplot2 is, I am wondering if this is possible through ggplot functions. I guess this involves using a specific stat (maybe with group=cyl?), but I am not able to.

Answer 1

We can plug your formula straight into stat_summary() to generate the desired result without intermediate steps:

library(ggplot2)
ggplot(mpg) + 
  stat_summary(aes(x = class, y = cyl), 
               fun.y = function(x) length(x) / length(unique(x)), 
               geom = "bar")