Manually change colour of one factor level in ggplot2 stacked barplot

Question

I'm giving a presentation, and would like to highlight the rate of occurrence of one of my factor levels on a slide by retaining the colour of this particular factor level, whilst greying out the other factor levels across other variables.

Using the Arthritis data set in the vcd library as an example:

library(vcd)
ggplot(Arthritis, aes(x = Sex, fill = Improved)) + geom_bar(position = "fill")

When, for example, I’m telling the audience about the marked improvement in the males, can I have a version of the plot which retains the colour of marked improvement in males, but greys out everything else? Preferably by using different shades of grey to keep the boundaries?

Answer 1

Consider this example:

library(scales)
library(ggplot2)

ggplot(Arthritis)+
  geom_bar(aes(x = Sex, group = Improved), fill = c("red", "blue", "green", "yellow", "grey50", "black"), position = "fill")

To change the specified field, while others keep the default ggplot colors:

ggplot(Arthritis)+
  geom_bar(aes(x = Sex,
               group = Improved),
           fill = c(rev(hue_pal()(3)),
                    "black",
                    hue_pal()(3)[2:1]),
            position = "fill")

or to grey out the rest with shades

ggplot(Arthritis)+
  geom_bar(aes(x = Sex,
               group = Improved),
           fill = c(paste0("grey", 7:9*10),
                    hue_pal()(3)[3],
                    paste0("grey", 8:9*10)),
           position = "fill")

to keep the legend just plot the above over your initial layer:

ggplot(Arthritis)+
  geom_bar(aes(x = Sex, fill = Improved),  position = "fill")+
  geom_bar(aes(x = Sex,
               group = Improved),
           fill = c(paste0("grey", 7:9*10),
                    hue_pal()(3)[3],
                    paste0("grey", 8:9*10)),
           position = "fill")

Answer 2

Another option is to replace the data points we'd like to highlight by NAs and then use scale_fill_grey to colour them with the argument na.value. We relabel the legend manually to show the original nomenclature.

Highlight all Marked

library(dplyr)    
ggplot(Arthritis %>% 
             mutate(Improved = replace(Improved, Improved == "Marked", NA))) + 
      geom_bar(position = "fill", aes(x = Sex, fill = Improved))+
      scale_fill_grey(start = 0.8, end = 0.6, na.value = "RoyalBlue", 
                      labels = c("None", "Some", "Marked"))

Highlight Male Marked

ggplot(Arthritis %>% 
         mutate(Improved = replace(Improved, Improved== "Marked" & Sex == "Male", NA))) + 
  geom_bar(position = "fill", aes(x = Sex, fill = Improved))+
  scale_fill_grey(start = 0.8, end = 0.2, na.value = "RoyalBlue",
                  labels = c("None", "Some", "Marked", "Marked"))

Upvotes: 1

Answer 3

You can add a column which contains the map of colors and use it with scales_fill_identity().

reprex::reprex_info()
#> Created by the reprex package v0.1.1.9000 on 2017-11-18

library(vcd)
#> Loading required package: grid

lvls <- levels(Arthritis$Improved)
color_maps <- scales::grey_pal(start = 0.8, end = 0.9)(length(lvls))
names(color_maps) <- lvls


library(dplyr, warn.conflicts = FALSE)

Arthritis_w_fill <- Arthritis %>%
  mutate(
    fill = if_else(
      Sex == "Male" & Improved == "Marked",
      "blue",
      color_maps[Improved]
  )
)


library(ggplot2)

ggplot(Arthritis_w_fill, aes(x = Sex, fill = fill)) +
  geom_bar(position = "fill") +
  scale_fill_identity(guide = "legend",
                      breaks = c(color_maps, "blue"),
                      labels = c(lvls, 'Sex == "Male" & Improved == "Marked"'))

Answer 4

Add

scale_fill_manual(values=c("Marked" = "RoyalBlue", "Some" = "DarkGrey", "None"="LightGrey"))

Full ggplot call:

ggplot(Arthritis, aes(x = Sex, fill = Improved)) + 
  geom_bar(position = "fill")+
  scale_fill_manual(values=c("Marked" = "RoyalBlue", "Some" = "DarkGrey", "None"="LightGrey"))