Percentage-y-scales with ..density.. in geom_histogram() works in facets with natural numbers, but not with (downscaled) decimal numbers

[ManuelNeumann]

I have a problem with percentage y-axes for histograms in facet-pots (using the scales-package). Everything worked just fine when I used the original scale, but it failed when I rescaled the items in an interval between 0 and 1.

Here's my example data:

# ----- Facet Plot ------
library(ggplot2)
library(scales)
library(magrittr)
library(dplyr)
library(reshape2)

# Trial 1 ----
set.seed(1234)
id <- c(1:40)

# Create five items with values between 0 and 5
x1 <- runif(n = 40, min = 0, max = 5) %>% round()
x2 <- runif(n = 40, min = 0, max = 5) %>% round()
x3 <- runif(n = 40, min = 0, max = 5) %>% round()
x4 <- runif(n = 40, min = 0, max = 5) %>% round()
x5 <- runif(n = 40, min = 0, max = 5) %>% round()

# Make data
data <- cbind(id, x1, x2, x3, x4, x5) %>% as.data.frame()

# Make Index
data %<>% mutate(x6 = (x1 + x2 + x3 + x4 + x5)/5)

Now I wanted to make a facet-plot, showing the histograms of each of the items and its respective percentage on the y-axis. Therefore I used aes(y = ..density..) (as I found out on Stack Overlow for the histograms and everything works perfectly fine. This was exactly the plot I needed:

# Long format data frame for facet plot
long.data <- melt(data, id.vars = "id")
head(long.data)

# And make the plot:
ggplot(data = long.data, aes(value)) + 
  geom_histogram(bins = 6,
                 aes(y = ..density..)) +
  facet_wrap(~variable) +
  scale_y_continuous(labels = percent_format()) +
  theme_bw()

# Everything is just fine: The continuous x-scale and a percentage-y-scale.

But then, I had to re-scale the items to the interval between 0 and 1 for different models and analyses:

# Trial 2 ----

data %<>% mutate(x1_resc = x1/6,
                 x2_resc = x2/6,
                 x3_resc = x3/6,
                 x4_resc = x4/6,
                 x5_resc = x5/6,
                 x6_resc = (x1_resc + x2_resc + x3_resc + x4_resc + x5_resc)/5)

As I ran the code again (the ggplot-code is identical to the one above), the y-axes of the histograms changed. Instead of the respective percentages, the axes show now percentage values between 0 and 200%.

# Long format
long.data <- melt(data[, c(1, 8:13)], id.vars = "id")
head(long.data)

# And the plot
ggplot(data = long.data, aes(value)) + 
  geom_histogram(bins = 6,
                 aes(y = ..density..)) +
  facet_wrap(~variable) +
  scale_y_continuous(labels = percent_format()) +
  theme_bw()

I can work with multiple workarounds (e.g. upscaling my items in the long-format and re-labelling the x-axis of the first plot), but I still hope this is an issue worth noticing.

[hadley]

That is because the y-axis is a density (the area integrates to 1), not a percentage (the heights sum to 1)

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/

[clauswilke]

Since this issue is getting some attention on twitter today: The solution is to use aes(y = stat(width*density)). This converts the density back into a percentage. Reprex follows below.

library(ggplot2)
library(scales)
library(magrittr)
library(dplyr)
library(reshape2)

# Trial 1 ----
set.seed(1234)
id <- c(1:40)

# Create five items with values between 0 and 5
x1 <- runif(n = 40, min = 0, max = 5) %>% round()
x2 <- runif(n = 40, min = 0, max = 5) %>% round()
x3 <- runif(n = 40, min = 0, max = 5) %>% round()
x4 <- runif(n = 40, min = 0, max = 5) %>% round()
x5 <- runif(n = 40, min = 0, max = 5) %>% round()

# Make data
data <- cbind(id, x1, x2, x3, x4, x5) %>% as.data.frame()

# Make Index
data %<>% mutate(x6 = (x1 + x2 + x3 + x4 + x5)/5)

# Long format data frame for facet plot
long.data <- melt(data, id.vars = "id")

# And make the plot:
ggplot(data = long.data, aes(value)) + 
  geom_histogram(bins = 6,
                 aes(y = stat(width*density))) +
  facet_wrap(~variable) +
  scale_y_continuous(labels = percent_format()) +
  theme_bw()

data %<>% mutate(x1_resc = x1/6,
                 x2_resc = x2/6,
                 x3_resc = x3/6,
                 x4_resc = x4/6,
                 x5_resc = x5/6,
                 x6_resc = (x1_resc + x2_resc + x3_resc + x4_resc + x5_resc)/5)

# Long format
long.data <- melt(data[, c(1, 8:13)], id.vars = "id")

# And the plot
ggplot(data = long.data, aes(value)) + 
  geom_histogram(bins = 6,
                 aes(y = stat(width*density))) +
  facet_wrap(~variable) +
  scale_y_continuous(labels = percent_format()) +
  theme_bw()

^{Created on 2018-12-19 by the reprex package (v0.2.1)}