Is there a way to create a ggplot facetted scatterplot with n colours, such that the colours alternate or are randomised from the palette?

Question

I've got 40 subjects in my dataset, 4 in each group and I want to create a plot that shows a line for each subject replicate (3 replicates); colouring them by subject, shaping by replicate. The problem I have is that the colours are so similar in each facet (group) that I can't really tell them apart.

My main body of code for the plot is:

ggplot(T_S, aes(x=as.numeric(Day), y=variable, colour=as.factor(Subj))) +
    geom_point(aes(shape=as.factor(Rep))) +
        geom_line(aes(linetype=as.factor(Rep))) +
            facet_wrap(.~Group,ncol=3) +
                 theme_bw() +
                    theme(legend.position="none")

And an example of what I mean by not being able to distinguish between colours is below using the viridis package. Is there a way to get the colours to alternate between the dark purple and yellow within each facet?

[Example with the Viridis Package][1]

Other things I've tried:

• scale_color_brewer(palette="Dark2")
• scale_fill_manual(values = wes_palette("GrandBudapest1", n = 38))
• scale_color_gradientn(colours = rainbow(40))

I also looked into the PolyChrome and randomcoloR packages, but can't see how they work with ggplot2. Any other suggestions also welcome! Thanks in advance for your help. [1]: https://i.stack.imgur.com/2iHXY.png

Answer 1

It's difficult for anyone to give you a tested, working solution on Stack Overflow if you don't include a sample of your data in the question. However, after a bit of reverse-engineering I have created what should be a reasonable approximation of your data structure:

set.seed(69)

T_S <- data.frame(Day      = rep(c(4, 11, 16, 25), 120),
                  Subj     = rep(1:40, each = 12),
                  Rep      = rep(rep(1:3, each = 4), 40),
                  Group    = factor(rep(1:10, each = 48), c(2:10, 1)),
                  variable = c(replicate(120, cumsum(runif(4, 0, 400)) + 250)))

And we can see that with your plotting code, we get similar results:

library(ggplot2)

ggplot(T_S, aes(x=as.numeric(Day), y=variable, colour=as.factor(Subj))) +
    geom_point(aes(shape=as.factor(Rep))) +
    geom_line(aes(linetype=as.factor(Rep))) +
    facet_wrap(.~Group,ncol=3) +
    theme_bw() +
    theme(legend.position="none") +
    scale_colour_viridis_d()

The reason for this is that your subject numbers are consecutive and correlate completely with the group. For the purposes of plotting, we want just four colors - one for each subject in each panel. The easiest way to achieve this is to renumber the subjects to 1:4 within each panel.

T_S$Subj <- T_S$Subj %% 4 + 1

So now the exact same plotting code gives us:

ggplot(T_S, aes(x=as.numeric(Day), y=variable, colour=as.factor(Subj))) +
    geom_point(aes(shape=as.factor(Rep))) +
    geom_line(aes(linetype=as.factor(Rep))) +
    facet_wrap(.~Group,ncol=3) +
    theme_bw() +
    theme(legend.position="none") +
    scale_colour_viridis_d()

Answer 2

It appears that it can be done through the package "RColorBrewer", by the combination of colours through different palettes.

Using the simulated data from Allan Cameron:

require(ggplot2)
require(RColorBrewer)
set.seed(69)
    
T_S <- data.frame(Day      = rep(c(4, 11, 16, 25), 120),
                  Subj     = rep(1:40, each = 12),
                  Rep      = rep(rep(1:3, each = 4), 40),
                  Group    = factor(rep(1:10, each = 48), c(2:10, 1)),
                  variable = c(replicate(120, cumsum(runif(4, 0, 400)) + 250)))

Now we can combine the colours from different palettes:

mycolours = c(brewer.pal(name="Accent", n = 7),
              brewer.pal(name="Dark2", n = 7),
              brewer.pal(name="Paired", n = 7),
              brewer.pal(name="Set3", n = 7),
              brewer.pal(name="Dark2", n = 7),
              brewer.pal(name="Set1", n = 7))

The palettes don't really matter in this instance, but I tried to choose them so they stood out. There is great documentation here for the full details on the colour palettes here: https://www.datanovia.com/en/blog/the-a-z-of-rcolorbrewer-palette/. More or fewer colours can be chosen as well, in this instance there are 42. Problems may occur if you need say, 1000 colours for instance, as there aren't that many in this package.

Finally the code to plot the data:

ggplot(T_S, aes(x=as.numeric(Day), y= variable, colour = as.factor(Subj))) +
  geom_point(aes(shape = as.factor(Rep))) + 
  geom_line(aes(linetype = as.factor(Rep))) +
  facet_wrap(.~Group, ncol = 3) +
  theme_bw()  +
  theme(legend.position="none") +
  scale_color_manual(values = mycolours)

https://i.stack.imgur.com/P07L5.jpg

GB