Is there an equivalent in ggplot to the varwidth option in plot?

Question

I am creating boxplots using ggplot and would like to represent the sample size contributing to each box. In the base plot function there is the varwidth option. Does it have an equivalent in ggplot?

For example, in base plot

data <- data.frame(rbind(cbind(rnorm(700, 0,10), rep("1",700)),
                         cbind(rnorm(50, 0,10), rep("2",50))))
data[ ,1] <- as.numeric(as.character(data[,1]))
plot(data[,1] ~ as.factor(data[,2]), varwidth = TRUE)

enter image description here

I seem to recall someone asking this on the mailing list quite a while ago and they were told it wasn't possible. I don't see anything referencing this in the issues on github, so it might still not be possible. (An alternative is to use fill colors.) — joran, Sep 28 '12 at 21:57
Not possible with ggplot, if you're only generating one plot you could possibly modify it in Illustrator or something similar — Omar Wagih, Sep 29 '12 at 04:32
@joran I have learnt from bitter experience that calling anything in R impossible just serves as bait for someone to prove you wrong. In this case the migthy @ kohske provided a workaround. — Andrie, Sep 29 '12 at 06:02
This has now been implemented with the `varwidth` argument. See this question: http://stackoverflow.com/q/25171210/3897439 — Cotton.Rockwood, Oct 20 '14 at 19:44

score 7 · Accepted Answer · answered Sep 29 '12 at 05:31

Not elegant but you can do that by:

data <- data.frame(rbind(cbind(rnorm(700, 0,10), rep("1",700)),
                         cbind(rnorm(50, 0,10), rep("2",50))))
data[ ,1] <- as.numeric(as.character(data[,1]))
w <- sqrt(table(data$X2)/nrow(data))
ggplot(NULL, aes(factor(X2), X1)) + 
  geom_boxplot(width = w[1], data = subset(data, X2 == 1)) +
  geom_boxplot(width = w[2], data = subset(data, X2 == 2))

enter image description here

If you have several levels for X2, then you can do without hardcoding all levels:

ggplot(NULL, aes(factor(X2), X1)) + 
  llply(unique(data$X2), function(i) geom_boxplot(width = w[i], data = subset(data, X2 == i)))

Also you can post a feature request: https://github.com/hadley/ggplot2/issues

score 2 · Answer 2 · answered Mar 02 '16 at 14:42

The current versions of ggplot2 (V 2.1.0) now contains a varwidth option:

data <- data.frame(rbind(cbind(rnorm(700, 0,10), rep("1",700)),
                     cbind(rnorm(50, 0,10), rep("2",50))))
data$X1 <- as.numeric(as.character(data$X1))
ggplot(data = data, aes(x = X2, y = X1)) + 
    geom_boxplot(varwidth = TRUE)

Is there an equivalent in ggplot to the varwidth option in plot?

2 Answers2

Linked

Related