geom_boxplot whisker length changing when changing y-axis scale

Question

The length of geom_boxplot() whisker shrinks when changing the y-axis scale and removing outliers. How can I make sure the whiskers do not shrink?

In the example below, the first plot shows all the data, inclusive of outliers. For cyl = 4, the upper whisker extends to mpg = 45. Similarly, the lower whisker for cyl = 8 extends to mpg = 11

In the second plot, the outliers are removed using outlier.shape = NA and the y-axis range is reduced. However, for cyl = 4 the upper whisker is now mpg = 34 and the lower whisker for cyl = 8 is now mpg = 13. I would expect the upper whisker for cyl = 4 to still be mpg = 45 and the lower whisker for cyl = 8 to still be mpg = 11.

library(ggplot2)
#> Warning: package 'ggplot2' was built under R version 4.2.3
library(dplyr)
#> Warning: package 'dplyr' was built under R version 4.2.3
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

# Load the mtcars dataset
data(mtcars)

data <- mtcars %>%
  select(mpg,cyl)

# Add outliers to specific groups
outliers <- data.frame(
  mpg = c(45,50,55,45,50,55),
  cyl = c(4,4,4,8,8,8))

# Create example dataset
data <- rbind(data, outliers)

# Upper whisker length for cyl = 4 is at 45 mpg
# Lower whisker length for cyl = 8 is at 11 mpg
data %>%
  ggplot(aes(x = factor(cyl), y = mpg)) +
  geom_boxplot()


# Remove outliers and reduce y-axis scale
# Upper whisker length for cyl = 4 is at 34 mpg
# Lower whisker length for cyl = 8 is at 13 mpg
data %>%
  ggplot(aes(x = factor(cyl), y = mpg)) +
  geom_boxplot(outlier.shape = NA) +
  scale_y_continuous(breaks = seq(0,45,5), limits = c(0,45))
#> Warning: Removed 4 rows containing non-finite values (`stat_boxplot()`).

^{Created on 2023-06-21 with reprex v2.0.2}

The whiskers are shortened due to you removing the outliers. This is a stats rule. By removing the outliers, you are focusing on the bulk of the data and eliminating those extreme values which distorts the visualization of the distribution — amanwebb, Jun 21 '23 at 15:14

score 2 · Accepted Answer · answered Jun 21 '23 at 15:22

By setting your limit to 45 you are actually removing any data outside of this range, resulting in a shorter plot.

You can tell by setting breaks = seq(0,55,5), limits = c(0,55) will keep your whisker ending at 45.

If you'd like to keep the plot between 0 and 45 you can use the function coord_cartesian(ylim=c(0, 45)) as in:

data %>%
  ggplot(aes(x = factor(cyl), y = mpg)) +
  geom_boxplot(outlier.shape = NA) +
  coord_cartesian(ylim=c(0, 45))

Please see Limit ggplot2 axes without removing data (outside limits): zoom for complete explanation.

geom_boxplot whisker length changing when changing y-axis scale

1 Answers1