CODEWITHSUNDEEP
rggplot2boxplotoutliers
DaniCee
DaniCee

Reputation: 3217

R ggplot2 - why does geom_boxplot ignore aesthetics "ymin", "lower", "middle", "upper", "ymax"?

For some testing purposes, I am trying to make boxplots where the upper and lower whiskers extend to the max and min data points (respectively), instead of treating them like outliers.

Not completely sure how this can be done best, but I figured I would just change my definition of upper and lower whisker to max() and min(), and pass them to geom_boxplot as ymin and ymax aesthetics (as indicated here: https://ggplot2.tidyverse.org/reference/geom_boxplot.html)

However, even when I get no warnings or errors, those aesthetics seem to be completely ignored.

See the MWE below with iris data:

data("iris")
iris_long <- as.data.frame(tidyr::pivot_longer(subset(iris, Species=="setosa"), -Species, names_to = "Var", values_to = "Value"))
iris_long_dt <- data.table::data.table(iris_long)
ddf <- as.data.frame(iris_long_dt[,list(y0=min(Value, na.rm=T),
                                        y100=max(Value, na.rm=T)),
                                  by=list(Var=Var)])
iris_long <- merge(iris_long, ddf, by="Var")
grDevices::png(filename="test.png", height=600, width=600)
P <- ggplot2::ggplot(iris_long, ggplot2::aes(x=Var, y=Value)) +
     ggplot2::geom_boxplot(ggplot2::aes(fill=Var, ymin=y0, ymax=y100),
                           position=ggplot2::position_dodge(.9)) +
     ggplot2::stat_summary(fun=mean, geom="point", shape=5, size=2) +
     ggplot2::theme_light()
print(P)
grDevices::dev.off()

This is what I get (which is identical to not passing ymin and ymax):

test

What I would expect is that the whiskers would extend to the min and max outlier data points (and hence the outliers not plotted, since they would not be outliers anymore).

Why is this happening? Am I doing something wrong? Thanks!

Upvotes: 0

Views: 496

Answers (1)

stefan
stefan

Reputation: 125398

Unfortunately it's not possible to provide only some of the boxplot stats. If you want to draw a boxplot manually you have to provide all stats and use stat="identity".

But to extend the whiskers over the whole data range you could use coef=Inf.

library(ggplot2)

ggplot(iris_long, aes(x = Var, y = Value)) +
  geom_boxplot(aes(fill = Var),
    position = position_dodge(.9), coef = Inf
  ) +
  stat_summary(fun = mean, geom = "point", shape = 5, size = 2) +
  theme_light()

enter image description here

Upvotes: 2

Related Questions