Independent y-axis per nested facet in ggplot

August 19, 2022

Is it possible to have a second "free" y-axis specifically for the nester "Short leaves"? I do not want to have an independent y-axis for all 3, just 2 for the respective nesters. How can I do that?

library(tidyverse)
library(ggh4x)

df <- as_tibble(iris) %>%
  select(3, 5) %>%
  mutate(Nester = if_else(Species == "setosa", "Short Leaves", "Long Leaves"),
         Nester = factor(Nester))

df %>%
  pivot_longer(!c(Species, Nester), names_to = "Measure", values_to = "Value") %>%
  ggplot(aes(Measure, Value)) +
  geom_boxplot() +
  facet_nested(~ Nester + Species)

EDIT:
So far, I only found those two options that do free y-axis for all 3:

facet_nested(~ Nester + Species, scales = "free_y", independent = "y")

and

facet_nested_wrap(~ Nester + Species, scales = "free_y", nrow = 1)

which do not give the desired result.

>Solution :

There isn’t really a good amount of control at the facet levels about which panel uses which y-scale. However, facetted_pos_scales() gives you exactly that control.

Let’s suppose we have this plot from your example.

library(tidyverse)
library(ggh4x)

df <- as_tibble(iris) %>%
  select(3, 5) %>%
  mutate(Nester = if_else(Species == "setosa", "Short Leaves", "Long Leaves"),
         Nester = factor(Nester))

p <- df %>%
  pivot_longer(!c(Species, Nester), names_to = "Measure", values_to = "Value") %>%
  ggplot(aes(Measure, Value)) +
  geom_boxplot() +
  facet_nested(~ Nester + Species, scales = "free_y", independent = "y")

We can ‘fix’ a scale by giving it a constant limit, which we’d have to pre-calculate. You can then set that scale to some panels.

ylim <- range(df$Petal.Length[df$Nester == "Long Leaves"])

p + facetted_pos_scales(
  y = list(Nester == "Long Leaves" ~ scale_y_continuous(limits = ylim))
)

If you also wish to omit the axis in between the panels with fixed scales, you’d need to set the scales separately for each panel. In the middle panel, you’d have to set guide = "none" to hide the axis.

p + facetted_pos_scales(
  y = list(
    Species == "versicolor" ~ scale_y_continuous(limits = ylim),
    Species == "virginica"  ~ scale_y_continuous(limits = ylim, guide = "none")
  )
)