0

I would like to produce a boxplot with wild outliers marked as a unique icon, say the asterisk #8. "Wild outliers" are defined as individual points that are more than Q3 + 3 * IQR or less than Q1 - 3 * IQR for the data set.

I have seen answers for people who want to label their outliers with their value (e.g. Labeling Outliers of Boxplots in R), and geom_boxplot() has built in a way to modify the style of all outliers. I haven't found any way to modify only some of the outlier points.

For this MRE, I would want to be able to have those wild outliers for 8 cylinders marked with an asterisk while the other outliers are marked with the usual filled in dot.

library(ggplot2)

ggplot(data = mtcars,aes(x=cyl,y=drat,group=cyl))+
  geom_boxplot()

Here is a way to get the "outer fences" beyond which points are considered wild outliers:

mtcars%>%group_by(cyl)%>%summarize(lf=quantile(drat,probs=.25)-3*IQR(drat),uf=quantile(drat,probs=.75)+3*IQR(drat))

Thank you!

Steven Ouellette
  • 67
  • 9

1 Answers1

1

One option would to create two separate dataframe containing the wild and the other outliers and add them to your boxplot via two geom_point.

library(ggplot2)
library(dplyr, warn = FALSE)

wild_outliers <- mtcars %>%
  group_by(cyl) %>%
  filter(drat < quantile(drat, probs = .25) - 3 * IQR(drat) |
    drat > quantile(drat, probs = .75) + 3 * IQR(drat))

outliers <- mtcars %>%
  group_by(cyl) %>%
  filter(drat < quantile(drat, probs = .25) - 1.5 * IQR(drat) |
    drat > quantile(drat, probs = .75) + 1.5 * IQR(drat)) |>
  anti_join(wild_outliers)
#> Joining, by = c("mpg", "cyl", "disp", "hp", "drat", "wt", "qsec", "vs", "am",
#> "gear", "carb")

ggplot(data = mtcars, aes(x = cyl, y = drat, group = cyl)) +
  geom_boxplot(outlier.colour = NA) +
  geom_point(data = outliers, shape = 16, size = 2) +
  geom_point(data = wild_outliers, shape = "*", size = 8)

stefan
  • 90,330
  • 6
  • 25
  • 51