ggplot2 histogram with ratios and facet_wrap

I am having issues with ggplot2 geom_histogram when plotting frequencies and using facet_wrap at the same time:

#
myTestDF<- data.frame(
  Sample = as.vector(replicate(n = 6, expr = c('s1', 's2', 's3'))),
  var2 = c(
    replicate(n = 9, expr = 't1'),
    replicate(n = 9, expr = 't2')),
  Val1 = c(
    replicate(n = 3, expr = c(2,20,40)),
    replicate(n = 3, expr = c(0.2,0.4,0.6))),
  stringsAsFactors = FALSE)
myTestDF<- rbind(myTestDF, data.frame(Sample = 's2', var2 = 't1', Val1 = 70,
                                      stringsAsFactors = FALSE)) ##afterthought :)
myTestDF$var3<- paste(myTestDF$Sample, myTestDF$var2, sep = '_')

###Now, this works:

ggplot(
  data = myTestDF[myTestDF$Sample=='s2',],
  aes(x = Val1, fill = var2)) +
  geom_histogram(
    aes(y = after_stat(c(
      count[group==1]/sum(count[group==1]),
      count[group==2]/sum(count[group==2])))),
    position = "identity", alpha = 0.5) +
  labs(title = 'testHist',
       x = "Val1", y = "Frequency") +
  theme_minimal() 

#

However, I can’t figure a way to make it work with facet_wrap by ‘Sample’. The frequencies get all messed up, and after experimenting and reading around, I can’t find a way to do it. Of course I can do a for loop , but I would like to understand if it can be done with facet_wrap or other ggplot2 function. Looking forward to your feedback.

>Solution :

The issue is that you do not account for the panels when computing the relative frequencies separately for each group, i.e. when you use facet_wrap the data is ordered first by PANEL and second by group. Instead I would suggest to use e.g. ave() to compute the relative frequencies. In the code below I also added the PANEL as a second grouping variable.

library(ggplot2)

ggplot(
  myTestDF, aes(x = Val1, fill = var2)
) +
  geom_histogram(
    aes(y = after_stat(
      ave(count, group, PANEL, FUN = \(x) x / sum(x))
    )),
    position = "identity", alpha = 0.5
  ) +
  labs(
    title = "testHist",
    x = "Val1", y = "Frequency"
  ) +
  theme_minimal() +
  facet_wrap(~Sample)
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.