How to create custom color palette to be used by scale_fill_manual()

Question

Consider the following code that makes a bar chart with a purple color palette

library(dplyr)
library(ggplot2)

dd <- mpg %>% 
        group_by(manufacturer, cyl) %>% 
        summarise(n = n()) %>%
        ungroup() 

mm <- dd %>%
        group_by(manufacturer) %>%
        summarise(mcyl = weighted.mean(cyl, n)) %>%
        arrange(mcyl) %>%
        ungroup()

dd %>% left_join(mm) %>%
        ggplot(mapping = aes(x = reorder(manufacturer, mcyl), y = n, fill = factor(cyl))) + 
        geom_bar(stat = "identity", position = "fill") +
        coord_flip() +
        scale_fill_brewer(palette = "Purples")

Question: How can I make the palette for Audi red ("Reds") and for Ford blue ("Blues"), while keeping the others purple ("Purples")?

What is the most convenient (preferably tidyverse) way to put these red/blue/purple palettes in a variable and passing it to scale_fill_manual() (as explained in this related Q&A)?

Sorry, not much time at the moment for a full answer. Make a colour palletes red, blue, purples, each 4, then map to car names with a new myCol column, use that column to fill? — zx8754, Aug 26 '16 at 12:34
I feel like this is generally fairly complicated. [This answer](http://stackoverflow.com/a/13016912/2461552) shows an example in a heatmap, but the approach may be similar for bars. — aosmith, Aug 26 '16 at 13:48
@TemplateRex You may also find [this Q&A](http://stackoverflow.com/questions/19568901/multiple-colour-scales-in-one-stacked-bar-plot-using-ggplot) relevant. — Henrik, Aug 26 '16 at 16:53
@Henrik follow-up question [here](http://stackoverflow.com/q/39173002/819272) — TemplateRex, Aug 26 '16 at 18:54

TemplateRex · Accepted Answer · 2016-08-30T18:25:00.970

Full working solution:

cyl <- sort(unique(mpg$cyl))
ncat <- length(cyl)          # 4 types of cylinders

# create palettes
library(RColorBrewer)
purples <- tibble(cyl, colr = brewer.pal(ncat, "Purples"))
reds    <- tibble(manufacturer = "audi", cyl, colr = brewer.pal(ncat, "Reds"))
blues   <- tibble(manufacturer = "ford", cyl, colr = brewer.pal(ncat, "Blues"))

# merge them with the data
dd_p <- dd %>% filter(!(manufacturer %in% c("audi", "ford"))) %>% left_join(purples)
dd_r <- dd %>% filter(manufacturer == "audi") %>% left_join(reds)
dd_b <- dd %>% filter(manufacturer == "ford") %>% left_join(blues)

gg_dd <- rbind(dd_p, dd_r, dd_b) %>%
        left_join(mm)

gg_dd %>% 
        ggplot(mapping = aes(x = reorder(manufacturer, mcyl), y = n, fill = colr)) + 
        geom_bar(stat = "identity", position = "fill") +
        coord_flip() +
        scale_fill_identity()

How to create custom color palette to be used by scale_fill_manual()

1 Answers1

Linked