Search code examples
rggplot2levels

Fixed color for characteristics of variable in barplot in R

I have a dataframe for which I want to visualize different things. In every barplot the same characteristic of the variable/attribute should be have the same color. For example:

data_mtcars <- mtcars

  data <- data_mtcars %>%                
    group_by(am, gear) %>%
    summarise(Freq = sum(mpg)) %>%
    group_by(am) %>%
    mutate(Prop = Freq / sum(Freq)) %>% 
    arrange(desc(Prop))

First plot with three characteristics of the variable "gear".

ggplot(data) +
  aes(x = am, y = Prop, fill = reorder(gear, Prop), width=0.5) +
  geom_col() + scale_y_continuous(labels = function(x) paste0(eval(x*100), "%")) +
  geom_text(aes(label = if_else(Prop>0.05, scales::percent(Prop),NULL)), position = position_stack(0.4))+    
  theme_minimal() +
  theme(legend.title = element_blank()) + ylab("") + xlab("") +
  scale_fill_brewer(palette = "Set3")

which gives me purple color for characteristic "3" of variable gear. If I change the number of characteristics the color should not be changed.

df <- data[data$gear!=4,]

ggplot(df) +
  aes(x = am, y = Prop, fill = reorder(gear, Prop), width=0.5) +
  geom_col() + scale_y_continuous(labels = function(x) paste0(eval(x*100), "%")) +
  # scale_x_date(breaks = unique(df_sum_EAD$Stichtag) , date_labels = "%d.%m.%Y") +
  geom_text(aes(label = if_else(Prop>0.05, scales::percent(Prop),NULL)), position = position_stack(0.4)) + theme_minimal() +
  theme(legend.title = element_blank()) + ylab("") + xlab("")+ scale_fill_brewer(palette = "Set3")

Now same characteristics have different colors (characteristic "3" is yellow). How to fix this problem. I have tried to fix the levels of the factor variable but I dont know how to include in an appropriate argument in plot.

data_mtcars$gear <- factor(data_mtcars$gear, levels=levels(as.factor(data_mtcars$gear)), ordered=T)
Solution

  • To get consistent colors for categories you can make use of a named color vector which can then be used via scale_color/fill_manual to always set the same color for each category:

    library(dplyr)
library(ggplot2)

data_mtcars <- mtcars

data <- data_mtcars %>%                
  group_by(am, gear) %>%
  summarise(Freq = sum(mpg)) %>%
  group_by(am) %>%
  mutate(Prop = Freq / sum(Freq)) %>% 
  arrange(desc(Prop))
#> `summarise()` regrouping output by 'am' (override with `.groups` argument)

data <- mutate(data, gear = reorder(gear, Prop))

# Named vector of colors
colors_gear <- scales::brewer_pal(palette = "Set2")(length(levels(data$gear)))
colors_gear <- setNames(colors_gear, levels(data$gear))

make_plot <- function(d) {
  ggplot(d) +
    aes(x = am, y = Prop, fill = reorder(gear, Prop), width=0.5) +
    geom_col() + scale_y_continuous(labels = function(x) paste0(eval(x*100), "%")) +
    geom_text(aes(label = if_else(Prop>0.05, scales::percent(Prop),NULL)), position = position_stack(0.4)) + theme_minimal() +
    theme(legend.title = element_blank()) + ylab("") + xlab("")+ 
    scale_fill_manual(values = colors_gear)
}

make_plot(data)


    make_plot(data[data$gear!=4,])