How to show seperate legend for two geom_col with same aestetics

September 6, 2023

I’m trying to plot overlapping stacked bar charts with the same aestetics and different widths of the bars.

My current solution is by adding separate geom_cols, but the same aestetics. Just different filters on the data set. But this gives the same fill colours two both. I can change this by adding alpha to one of them, which looks OK for me.

The problem here is then, that I still get only one legend for both bars as you can see in the following example:

library(tidyverse)

data <- tribble(
  ~status , ~category , ~month , ~value ,
  'plan' , 'A' , 'Jan' , 5000 ,
  'plan' , 'A' , 'Feb' , 4900 ,
  'plan' , 'A' , 'Mar' , 4900 ,
  'plan' , 'B' , 'Jan' , 0 ,
  'plan' , 'B' , 'Feb' , 499 ,
  'plan' , 'B' , 'Mar' , 600 ,
  'actual' , 'A' , 'Jan' , 5200 ,
  'actual' , 'A' , 'Feb' , 4950 ,
  'actual' , 'A' , 'Mar' , 4700 ,
  'actual' , 'B' , 'Jan' , 0 ,
  'actual' , 'B' , 'Feb' , 380 ,
  'actual' , 'B' , 'Mar' , 400
) |>
  mutate(month = factor(month , levels = month.abb[1:3]))

data |>
  ggplot(mapping = ggplot2::aes(x = month , y = value)) +
  geom_col(data = ~ . |> filter(status == 'plan') ,
                    mapping = ggplot2::aes(fill = category) ,
                    alpha = 0.5 ,
                    position = position_stack(reverse = TRUE)) +
  geom_col(data = ~ . |> filter(status == 'actual') ,
                    mapping = aes(fill = category) ,
                    position = position_stack(reverse = TRUE) ,
                    width = 0.5) +
  scale_fill_manual(name = 'Category' ,
                             values = c('A' = '#628395' , 'B' = '#95A472') ,
                             labels = c('A' = 'one' , 'B' = 'two'))

I can put alpha in the aes() of both bars and add a scale_alpha_manual, this gives an additional legend, but with wrong colours.

So, does anyone have an idea, how I can achieve a similar plot with two legends?

>Solution :

You can use fill= on an "interaction" of status and category.

data |>
  ggplot(mapping = ggplot2::aes(x = month , y = value)) +
  geom_col(data = ~ . |> filter(status == 'plan') ,
                    mapping = ggplot2::aes(fill = interaction(status, category)) ,
                    position = position_stack(reverse = TRUE)) +
  geom_col(data = ~ . |> filter(status == 'actual') ,
                    mapping = aes(fill = interaction(status, category)) ,
                    position = position_stack(reverse = TRUE) ,
                    width = 0.5) +
  scale_fill_manual(name = 'Category' ,
                    values = c('actual.A' = '#628395' , 'actual.B' = '#95A472',
                               'plan.A' = '#62839588' , 'plan.B' = '#95A47288') ,
                    labels = c('A' = 'one' , 'B' = 'two'))

The labels actual.A and such may not be inspiring. All interaction is doing is creating a concatenated string collapsed with ., so you can create your own new field that combines category and status in a way that you prefer.