Adding subcategories to a legend in ggplot2

Question

Let's say we have this following hierarcichal data on the habitats that make up my fantasy island (which is of course always warm and sunny!)

set.seed(1)

hab_dat <- data.frame(
  habitat_type = rep(c("sea", "coast", "land"), times = 1, each = 3),
  habitat_name = c("rocky", "sandy", "seaweed",
                  "beach", "pebbles", "rockpools",
                  "fields", "hills", "forest"),
  area_km2 = sample(10:40, size =9))
  
hab_dat

I want to plot the total area of each habitat type and so write following code

hab_dat %>% 
  group_by(habitat_type) %>% 
  summarise(area_km2 = sum(area_km2)) %>%
  ggplot(aes(x = habitat_type, y = area_km2, fill = habitat_type)) +
  geom_bar(stat = "identity") +
  scale_fill_manual(values = c("gold", "forestgreen", "blue"))

Looks good, but the legend is not very informative. I would like for the habitats contained within each habitat type to be included in the legend under the appropriate habitat type, just as qualitative information. Here is an example I made in paint.

I can get a bit closer using the following code without affecting the appearance of the plot, however, I am missing the habitat_type titles and also have multiple tiles for the same colour.

hab_dat <- hab_dat %>% mutate(col = rep(c("blue", "gold", "forestgreen"), times = 1, each = 3))

pal <- setNames(as.character(hab_dat$col), as.character(hab_dat$habitat_name))

ggplot(hab_dat, aes(x = habitat_type, y = area_km2, fill = habitat_name)) +
  geom_bar(position = "stack", stat = "identity") +
  scale_fill_manual(values = pal)

I have been looking at solutions along the lines of this one but am trying for a more automated solution as my actual data is a bit larger than this, and also one that presents the colour tile once per group as per my drawing.

I am also open to solutions that involve creating the legend separately outside of ggplot. — rainbird, Jul 19 '22 at 13:57

score 2 · Accepted Answer · answered Jul 19 '22 at 13:48

I don't think there is an elegant solution that adresses your problem. I'll suggest here that you format the labels to imply the hierarchy.

library(ggplot2)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

set.seed(1)

hab_dat <- data.frame(
  habitat_type = rep(c("sea", "coast", "land"), times = 1, each = 3),
  habitat_name = c("rocky", "sandy", "seaweed",
                   "beach", "pebbles", "rockpools",
                   "fields", "hills", "forest"),
  area_km2 = sample(10:40, size =9))

# Format labels
labels <- split(hab_dat$habitat_name, hab_dat$habitat_type)
labels <- unlist(Map(function(top, bottom) {
  paste0(top, "\n", paste("- ", bottom, collapse = "\n"))
}, top = names(labels), bottom = labels))


hab_dat %>% 
  group_by(habitat_type) %>% 
  summarise(area_km2 = sum(area_km2)) %>%
  ggplot(aes(x = habitat_type, y = area_km2, fill = habitat_type)) +
  geom_bar(stat = "identity") +
  scale_fill_manual(
    values = c("gold", "forestgreen", "blue"),
    labels = function(i) {labels[i]} # Lookup label
  )

^{Created on 2022-07-19 by the reprex package (v2.0.1)}

@rainbird can you explain why it is only a step closer and not a complete solution? It's difficult to see what needs to be improved in this answer to meet your specification. — Allan Cameron, Jul 19 '22 at 14:00
I am trying to arrange it so that the colour tile is only next to only the habitat type. This may seem like I am being picky, but in a version of the plot with different numbers of sub categories the colour tiles in the legends appear as different lengths. — rainbird, Jul 19 '22 at 14:13
Thanks again for this @teunbrand. Accepted as the closest answer possible and works great. — rainbird, Jul 21 '22 at 10:12

Adding subcategories to a legend in ggplot2

1 Answers1