ggplot: plot 4th discrete variable respective of 3rd continuous variable-CodePudding

I have a dataset that looks like this:

d <- data.frame(w = runif(3000),
                x = runif(3000),
                y = rnorm(3000),
                z = rep(LETTERS[1:3], 1000))

I can plot w as a continuous variable with some colour palette; scale_colour_viridis_c() does this nicely:

library(ggplot2)
ggplot(d, aes(x, y))  
 geom_point(aes(colour = w))  
 scale_colour_viridis_c()

But say I want to express w in terms of z too. I can't with the option argument in scale_colour_viridis_c() because "EXPR must be a length 1 vector".

How can I do this such that there are more than one colour palette used (i.e., unique(d$z))?

CodePudding user response：

One option to have multiple color (or ...) scales for the same aesthetic would be the ggnewscale package which TBMK requires to split your data by z and use multiple layers. To make this a bit easier I use purrr::imap to create the layers for each z:

set.seed(123)

library(ggplot2)
library(ggnewscale)

d_split <- split(d, d$z)

limits <- range(d$w)

ggplot(mapping = aes(x, y))  
  purrr::imap(d_split, function(x, y) {
    list(
      geom_point(aes(colour = w), data = x),
      scale_colour_viridis_c(option = y, limits = limits),
      new_scale_colour()
    )
  })

CodePudding user response：

Modifying your sample data slightly so that the plot is at least vaguely sensible:

library(tidyverse)
library(ggplot2)

d <- data.frame(w = round(runif(99), 1),
                x = runif(99),
                y = round(rnorm(99), 1),
                z = rep(LETTERS[1:3], 33))

ggplot(d, aes(x, y))  
  geom_point(aes(colour = interaction(w, z)))

interaction() creates a discrete categorisation so you can't use scale_colour_viridis_c().