ggplot2: Qualitative Colour Palettes

Every now and then I would like to change the colours of my plots. ggplot2 provides several different ways to accomplish this task, however choosing the one that matches my mood on that particular day is always relatively time-consuming. So I decided to make a post highlighting some of the available colour palette options. In this post I will focus on the colour schemes helping to depict categorical data, and will try to use various predetermined colour palettes as much as possible, as I find hard to distinguish between the intricacies of different color spaces including RGB, HSV, HLS, CIEXYZ, CIELUV, HCL (polar CIELUV), CIELAB and polar CIELAB.

Default formatting

First, load the necessary libraries and set up a dummy dataframe of 8 variables for plotting. As the main focus is on the colours I remove the legends, titles and ticks from the plots.

library(ggplot2)
library(colorspace)
df <- data.frame(x=letters[1:8])p <- qplot(data=df, x, 1, geom="bar", stat="identity", fill=x) +
opts(
legend.position = "none",
axis.title.x = theme_blank(),
axis.title.y = theme_blank(),
axis.text.x = theme_blank(),
axis.text.y = theme_blank(),
axis.ticks = theme_blank(),
title="Default colours")

Colorbrewer

ColorBrewer is an “online tool designed to help people select good color schemes for maps and other graphics”. ggplot2 supports the use of their colour schemes out-of-the-box.

brewerplot <- function (palette) {
p + scale_fill_brewer(palette = palette) + opts(title=palette)
}
brewerplot ("Set1")

brewerplot ("Set2")

brewerplot ("Set3")

brewerplot ("Accent")

brewerplot ("Dark2")

brewerplot ("Paired")

brewerplot ("Pastel1")

brewerplot ("Pastel2")

Colorspace

R package colorspace includes several palettes based on HCL colours. The plots below are using the palettes for 4 categorical variables given in the colorspace documentation.

colours <- rainbow_hcl(4, start = 30, end = 300)
p %+% df2 + scale_fill_manual (values=colours) + opts(title="dynamic")

colours <- rainbow_hcl(4, start = 60, end = 240)
p %+% df2 + scale_fill_manual (values=colours) + opts(title="harmonic")

colours <- rainbow_hcl(4, start = 270, end = 150)
p %+% df2 + scale_fill_manual (values=colours) + opts(title="cold")

colours <- rainbow_hcl(4, start = 90, end = -30)
p %+% df2 + scale_fill_manual (values=colours) + opts(title="warm")

Colorspace using scale_fill_hue

It is also possible to use the colorspace palettes directly through scale_fill_hue(). However, ggplot2 version 0.8.2 produces somewhat different colours. Hadley explains: “scale_fill_hue generates n + 1 colours between start and end and then drops the last colour. That’s the best method I could come up with for when you have n colours spanning 0 to 360, but it’s not quite right when you restrict the range of the hues”. In the upcoming version this behaviour has been changed and I will be updating this post once the version update becomes available.

Version 0.8.3 improved the scale_fill_hue() function and the plots below are if not identical then at least very similar to the ones created with the help of colorspace package.
p %+% df2 + scale_fill_hue(c = 50, l = 70, h=c(30, 300)) + opts(title="scale_fill_hue-dynamic")

p %+% df2 + scale_fill_hue(c = 50, l = 70, h=c(60, 240)) + opts(title="scale_fill_hue-harmonic")

p %+% df2 + scale_fill_hue(c = 50, l = 70, h=c(270, 150)) + opts(title="scale_fill_hue-cold")

p %+% df2 + scale_fill_hue(c = 50, l = 70, h=c(90, -30)) + opts(title="scale_fill_hue-warm")

Custom Palette

If you need to use a specific colour palette (e.g. colours specified in your corporate brand manual), then this can be accomplished using scale_fill_manual(). There are several websites (ColorBlender, Kuler, COLOURlovers, to name just a few) which allow you to create your own palettes, or to use palettes created by other users. In the example below, Primary Peter 5 palette from ColorBlender.com is being used.

colours <- c("#2121D9", "#9999FF", "#D92121", "#21D921", "#FFFF4D", "#FF9326")
p + scale_fill_manual (values=colours) + opts(title="Primary Peter 5")