boxplot: order groups by the mean of a subset of each group

Question

Let's consider this data:

df = data.frame('score'=round(runif(15, 1, 10)),
                'group'=paste0("a",rep(c(1,2,3),each=5)),
                'category'=rep(c("big", "big", "big", "big", "small"), 3))

I would like to plot boxplots of this data with ggplot2. What i want is: boxplot(score~group), but with the boxplots arranged according to the mean of the "big" individuals of each group.

I can't figure it out in a simple way, without creating new variables. OK to use Dplyr. Thanks.

Answer 1

I don't know if this qualifies as a simple way, I personally find it simple, but I use dplyr to find the means:

#find the means for each group
library(dplyr)
means <-
df %>%
  #filter out small since you only need category equal to 'big'
  filter(category=='big') %>%
  #use the same groups as in the ggplot
  group_by(group) %>%
  #calculate the means
  summarise(mean = mean(score))

#order the groups according to the order of the means
myorder <- means$group[order(means$mean)]

In this case the order is:

> myorder
[1] a1 a2 a3

In order to arrange the order of the boxplots according to the above you just need to do:

library(ggplot2)
ggplot(df, aes(group, score)) +
  geom_boxplot() +
  #you just need to use scale_x_discrete with the limits argument
  #to pass in details of the order of appearance for the boxplots
  #in this case the order is the myorders vector
  scale_x_discrete(limits=myorder)

And that's it.