rggplot2kernel-densityviolin-plot

Separating groups in plot when using gradient fill

I'm using stat_ydensity (co-opted from the violin plot) to shade bar density using color rather than size.

This code does what I need, when using a single category:

library(tidyverse)
set.seed(100)
n <- 100
df <- data.frame(a=rnorm(n, mean=5, sd=1), b=rnorm(n, mean=10, sd=2), c=rnorm(n, mean=15, sd=3)) %>%
  pivot_longer(everything(), names_to = "group", values_to = "value") %>%
  mutate(group = factor(group),
         type = as.character(rbinom(seq(nrow(.)), 1, 0.5)))

ggplot() + 
  stat_ydensity(data = df,
                aes(x = group,
                    y = value,
                    fill = after_stat(density), 
                    xmin = stat(x) - 0.2, xmax = stat(x) + 0.2,
                    # Nudge y by a bit depending on the spread of your data
                    ymin = stat(y) - 0.05, ymax = stat(y) + 0.05), 
                geom = "rect", trim = TRUE)

enter image description here

However, I'd like to be able to compare distributions in two different categories. Normally, this would be done using aes(fill) or aes(color):

ggplot() + 
  stat_ydensity(data = df,
                aes(x = group,
                    y = value,
                    fill = type,
                    xmin = stat(x) - 0.2, xmax = stat(x) + 0.2,
                    ymin = stat(y) - 0.05, ymax = stat(y) + 0.05), 
                geom = "rect", trim = TRUE)

But this isn't compatible with the using fill to do density. Is there any way to replicate the first plot, but while having two groups in separate colors?

Solution

  • Here is a sketch of a possible solution using the package ggnewscale.

    
library(tidyverse)
library(ggnewscale)
set.seed(100)
n <- 200
df <- data.frame(a=rnorm(n, mean=5, sd=1), b=rnorm(n, mean=10, sd=2), 
                 c=rnorm(n, mean=15, sd=3)) %>%
  pivot_longer(everything(), names_to = "group", values_to = "value") %>%
  mutate(group = factor(group),
         type = as.character(rbinom(seq(nrow(.)), 1, 0.5)))



ggplot() + 
  stat_ydensity(data = filter(df, type == 0),
                aes(x = as.numeric(group) - 0.2,
                    y = value,
                    group = interaction(group,type),
                    fill = after_stat(density), 
                    xmin = stat(x) - 0.2, xmax = stat(x) + 0.2,
                    # Nudge y by a bit depending on the spread of your data
                    ymin = stat(y) - 0.05, ymax = stat(y) + 0.05), 
                inherit.aes = FALSE,
                geom = "rect", trim = TRUE) +
  scale_fill_continuous(name = "type 0") +
  new_scale_fill() +
  stat_ydensity(data = filter(df, type ==1),
                aes(x = as.numeric(group) +0.2,
                    y = value,
                    group = interaction(group,type),
                    fill = after_stat(density), 
                    xmin = stat(x) - 0.2, xmax = stat(x) + 0.2,
                    # Nudge y by a bit depending on the spread of your data
                    ymin = stat(y) - 0.05, ymax = stat(y) + 0.05), 
                inherit.aes = FALSE,
                geom = "rect", trim = TRUE) +
  scale_fill_gradient(low = 'blue', high = 'red', name = "type 1") +
  scale_x_continuous(breaks = c(1,2,3), labels = c("a","b","c"),name = 'group')
#> Warning: `stat(x)` was deprecated in ggplot2 3.4.0.
#> ℹ Please use `after_stat(x)` instead.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.

<sup>Created on 2025-06-15 with [reprex v2.1.1](https://reprex.tidyverse.org)</sup>

    enter image description here