Boxplot alternatives

Visualization Best Practices in R

Nick Strayer

Instructor

Limitations of the boxplot with jitter

• Jostling points can only deal with so much overlap
• Hard to get an idea of data density

What are some other options?

Beeswarm plots

Violin plots

Beeswarm plots

• 'Smart' jittering
• Individual points are clumped together as close to the axis as possible
• Handily included as geom_beeswarm() in the ggbeeswarm package.

library(ggbeeswarm)
ggplot(data, aes(y = y, x =  group)) + 
  geom_beeswarm(color = 'steelblue')

Beeswarm pros

• Individual data points
• Distributional shape

Beeswarm cons

• Get hard with lots of data
• Arbitrary stacking

Violin plots

• KDE reflected to be symmetric
• Just replace geom_boxplot() with geom_violin().

ggplot(data, aes(y = y, x = group)) + 
  geom_violin(fill = 'steelblue') 

Violin pros

• Every data point is heard
• Not every data point is seen, so good for lots of data.

Violin cons

• Kernel width choice
• Not every data point is seen

Let's try some more advanced comparisons!

Visualization Best Practices in R