Efficient way to produce facetted bar plots

Question

Is there a way to produce facetted bar plots as rapidly as you can produce facetted box plots?

For instance this code will produce box plots facetted by a grouping variable:

library(tidyverse)

mtcars %>% 
  ggplot(aes(x = am, y = mpg)) + 
  geom_boxplot() + 
  facet_grid(~ cyl)

But the "same" code for a bar plot fails because geom_col() expects to have means provided, which would take considerably more time to calculate:

mtcars %>% 
  ggplot(aes(x = am, y = mpg)) + 
  geom_col() + 
  facet_grid(~ cyl)

If there is a way to produce facetted bar plots at the speed of facetted box plots, I'd love to hear your methods.

Answer 1

There seems to be no notable difference between those two plots

library(tidyverse)

box_plot <- function() mtcars %>% 
  ggplot(aes(x = am, y = mpg)) + 
  geom_boxplot() + 
  facet_grid(~ cyl)

bar_plot <- function() mtcars %>% 
  ggplot(aes(x = am, y = mpg)) + 
  geom_col() + 
  facet_grid(~ cyl)

library(microbenchmark)

mb <- microbenchmark(box_plot(), bar_plot())

mb
# Unit: milliseconds
#        expr      min       lq     mean   median       uq      max neval cld
#  box_plot() 2.412589 2.431660 2.538586 2.475522 2.534515 9.966422  1000   a 
#  bar_plot() 2.543288 2.566173 2.670544 2.609146 2.671571 5.254653  1000   b

autoplot(mb)

A similar test with the full diamonds dataset yields the same result. I think the bottleneck is the graphical device that needs a lot of time to display a ggplot object.

box_p <- box_plot()
bar_p <- bar_plot()

# !!! this will need about 2 minutes to execute !!!
mb2 <- microbenchmark(plot(box_p), plot(bar_p), times = 100)

mb2
# Unit: milliseconds
#         expr      min       lq     mean   median       uq      max neval cld
#  plot(box_p) 405.0375 513.7360 532.1827 524.1259 536.8206 740.7494   100   b
#  plot(bar_p) 454.6047 469.3411 489.3676 479.3709 497.9246 717.7909   100  a 

autoplot(mb2)

Answer 2

Thank you to @hrbrmstr for your suggestion to use stat_summary() . That was the perfect solution.

mtcars %>% 
  ggplot(aes(x = as.factor(am), y = mpg)) + 
  stat_summary(fun.y = "mean", geom = "col") + 
  facet_grid(~ cyl)