How to order stacked bar plot x categories by value of one of the fill categories in ggplot2/R

Question

I'm trying to plot my data as a stacked bar with 3 levels ("catg"), but I want the categories on the X-axis to appeared in increasing order by their value of the "low" sub-categories,

here is reproductive example:

#creating df:
set.seed(33)
df<-data.frame(value=runif(12),
               catg=factor(rep(c("high","medium","low")),
                           levels = c("high","medium","low")),
               var_name=(c(rep("question1",3),rep("question2",3),rep("question3",3),rep("question4",3)))
#plotting
bar_dist<-ggplot(df,aes(x=var_name,
                              y=value, 
                              fill=catg, 
                             label=round(value,2)))

bar_dist+ geom_bar(stat = "identity",
                   position = "dodge", 
                   width = 0.7)+
  coord_flip() +
  xlab("questions")+
  ylab("y")+
  geom_text(size = 4,position=position_dodge(width = 0.7))

And here is my current plot:

so in this case I should have question3 and then 4, 1, and finally 2. every help will be appreciate,

Answer 1

A solution that doesn't modify the data frame, using fct_reorder2() from the forcats library:

library(forcats)

bar_dist <- ggplot(df,
                   aes(
                     x = fct_reorder2(var_name, catg, value),
                     y = value, fill = catg, 
                     label = round(value, 2)))

bar_dist + geom_bar(stat = "identity",
                   position = "dodge", 
                   width = 0.7) +
  coord_flip() +
  xlab("questions") +
  ylab("y") +
  geom_text(size = 4, position = position_dodge(width = 0.7))

Answer 2

One way could be :

df$var_name=factor(df$var_name,levels=rev(levels(reorder(df[df$catg=="low",]$var_name,df[df$catg=="low",]$value))))

It uses reorder() as suggested by Richard Telford to reorder the levels according to df$value after filtering df to keep only the "low" .
levels() is used to extract the levels from the previous function.
rev() is used to reverse the levels.
factor() reassigns the levels to df$var_name

or :

df$var_name=factor(df$var_name,levels = df[with(df,order(value,decreasing = T)) ,][df[with(df,order(value,decreasing = T)) ,]$catg=="low",]$var_name)

It sorts df on df$value (by decreasing value), filters on df$catg for "low" and retrieves df$var_name which is used as levels in factor() .

The same plotting function is then used: