4

I have a boxplot that I made in R with ggplot2 analagous to the sample boxplot below.

enter image description here

The problem is, for the values on the y axis (in this sample, the number of cylinders in the car) I have very different frequencies -- I may have included 2 8 cylinder cars, but 200 4 cylinder cars. Because of this, I'd like to be able to resize the boxplots (in this case, change the height along the y axis) so that the 4 cylinder boxplot is a larger portion of the chart than the 8 cylinder boxplot. Does someone know how to do this?

MrFlick
  • 195,160
  • 17
  • 277
  • 295
Dkarode
  • 123
  • 1
  • 2
  • 5
  • 1
    I think you want `varwidth` - see the help page for `geom_boxplot`. – aosmith Aug 06 '14 at 22:44
  • By the way, you should post code for reproducibility... it makes it much easier for people to answer your question. See "how to make a great R reproducible example" http://stackoverflow.com/q/5963269/3897439 – Cotton.Rockwood Aug 07 '14 at 02:24
  • Add `geom_point` (or `geom_jitter`), set it to a bit of transparent and show your data. At that point, one could even omit a boxplot. – Roman Luštrik Aug 07 '14 at 17:24

1 Answers1

4

As @aosmith mentioned, varwidth is the argument you want. It looks like it may have been accidentally removed from ggplot2 at some point (https://github.com/hadley/ggplot2/blob/master/R/geom-boxplot.r). If you look at the commit title, it is adding back in the varwidth parmeter. I'm not sure if that ever made into the cran package, but you might want to check your version. It works with my version: ggplot2 v.1.0.0 I'm not sure how recently the feature was added.

Here is an example:

library(ggplot2)

set.seed(1234)
df <- data.frame(cond = factor( c(rep("A",200), rep("B",150), rep("C",200), rep("D",10)) ), 
                 rating = c(rnorm(200),rnorm(150, mean=0.2), rnorm(200, mean=.8), rnorm(10, mean=0.6)))

head(df, 5)
tail(df, 5)

p <- ggplot(df, aes(x=cond, y=rating, fill=cond)) + 
  guides(fill=FALSE) + coord_flip()

p + geom_boxplot()

Gives: enter image description here

p + geom_boxplot(varwidth=T)

Gives: enter image description here

For a couple of more options, you can also use a violin plot with scaled widths (the scale="count" argument):

p+ geom_violin(scale="count")

enter image description here

Or combine violin and boxplots to maximize your information.

p+ geom_violin(scale="count") + geom_boxplot(fill="white", width=0.2, alpha=0.3)

enter image description here

Cotton.Rockwood
  • 1,601
  • 12
  • 29