How do you create a gradient of colors for a discrete variable in ggplot2?

19
votes

I have data with about 100 ordered categories. I would like to plot each category as a line separately, with the line colors ranging from a low value (say, blue) to a high value (say, red).

Here's some sample data, and a plot.

# Example data: normal CDFs

library(ggplot2)

category <- 1:100
X <- seq(0, 1, by = .1)
df <- data.frame(expand.grid(category, X))
names(df) <- c("category", "X")
df <- within(df, {
  Y <- pnorm(X, mean = category / 100)
  category <- factor(category)
  })

# Plot with ggplot
qplot(data = df, x = X, y = Y, color = category, geom = "line")

This produces a pretty rainbow thing (below)enter image description here

but I'd rather have a gradient from blue to red. Any ideas how I can do that?

2
rggplot2

2 Answers

25
votes

The default gradient functions for ggplot expect a continuous scale. The easiest work around is to convert to continuous like @Roland suggested. You can also specify whatever color scale you want with scale_color_manual. You can get the list of colors ggplot would have used with

cc <- scales::seq_gradient_pal("blue", "red", "Lab")(seq(0,1,length.out=100))

This returns 100 colors from blue to red. You can then use them in your plot with

qplot(data = df, x = X, y = Y, color = category, geom = "line") +     
    scale_colour_manual(values=cc)

enter image description here

14
votes

Since a discrete legend is useless anyway, you could use a continuous color scale:

ggplot(data = df, aes(x = X, y = Y, color = as.integer(category), group = category)) +
  geom_line() +
  scale_colour_gradient(name = "category", 
                        low = "blue", high = "red")

resulting plot