Allow users to input their own colour schemes, fixes for problems with >9 groups.

[quantixed]

This PR is an enhancement to {dabestr} and fixes a couple of issues with colouring dabest_plots.

• Enhancement: Users should be able to supply a custom palette for use in graphics (issue Customizing Color Palettes in dabest_plot #160)
• Fix: if a dabest_plot with more than 9 groups is requested, resulting plot is missing bars (issue This manual palette can handle a maximum of 10 values. You have supplied 12  #175)
• Fix: if a dabest_plot with more than 9 groups is requested, colour scheme is broken (issue This manual palette can handle a maximum of 10 values. You have supplied 12  #175)
• Fix: If a dabest_plot with more than n groups is requested, plot creation fails (where n is the maximum colours in custom palette, issue This manual palette can handle a maximum of 10 values. You have supplied 12  #175).

I also merged the changes from marsiwiec which fixes the formatting of the print statements and adds an option to silence the greeting.

Issues (MWE)

Using the current version of {dabestr} the following code illustrates the issues:

library(dabestr)
set.seed(12345)
N <- 20

df <- tibble::tibble(
  ca = rnorm(N, mean = 3, sd = 0.4),
  cb = rnorm(N, mean = 3.5, sd = 0.75),
  cc = rnorm(N, mean = 3.25, sd = 0.4),
  cd = rnorm(N, mean = 3, sd = 0.4),
  ce = rnorm(N, mean = 3.5, sd = 0.75),
  cf = rnorm(N, mean = 3.25, sd = 0.4),
  cg = rnorm(N, mean = 3, sd = 0.4),
  ch = rnorm(N, mean = 3.5, sd = 0.75),
  ci = rnorm(N, mean = 3.25, sd = 0.4),
  cj = rnorm(N, mean = 3.5, sd = 0.75),
  ta = rnorm(N, mean = 3.5, sd = 0.5)
)

df <- df |>
  tidyr::gather(key = Group, value = Measurement)

sub9.mean_diff <- load(df, x = Group, y = Measurement,
                       idx = c(
                         "ca", "cb", "cc", "cd", "ce",
                         "cf", "cg", "ch", "ci"
                       )) |> mean_diff()

sub10.mean_diff <- load(df, x = Group, y = Measurement,
                        idx = c(
                          "ca", "cb", "cc", "cd", "ce",
                          "cf", "cg", "ch", "ci", "cj"
                        )) |> mean_diff()

sub11.mean_diff <- load(df, x = Group, y = Measurement,
                        idx = c(
                          "ca", "cb", "cc", "cd", "ce",
                          "cf", "cg", "ch", "ci", "cj", "ta"
                        )) |> mean_diff()

# jama has only 7 colours - requested plot fails with empty groups for 8 & 9
dabest_plot(sub9.mean_diff, custom_palette = "jama")

# output is ok using uscgb, which has enough colours. 9 groups works as intended
dabest_plot(sub9.mean_diff, custom_palette = "ucscgb")

# output is in incorrect with 10 groups
# - colours do not match between raw and delta plots
dabest_plot(sub10.mean_diff, custom_palette = "ucscgb")

# output is in incorrect with 11 groups
# - colours do not match between raw and delta plots
# - colour mismatch between swarm plot and swarm bars in raw plot
dabest_plot(sub11.mean_diff, custom_palette = "ucscgb")

With the proposed changes, the last four commands all work correctly.

The enhancement I am proposing is to add manual colouring to the graphics.

# ENHANCEMENT: Define a custom palette with enough colors
my_palette <- c("#e6194b", "#3cb44b", "#ffe119", "#4363d8", "#f58231",
                "#911eb4", "#46f0f0", "#f032e6", "#bcf60c", "#fabebe",
                "#008080", "#e6beff", "#9a6324", "#fffac8", "#800000",
                "#aaffc3", "#808000", "#ffd8b1", "#000075", "#808080",
                "#ffffff", "#000000")

dabest_plot(sub11.mean_diff,
            custom_palette = "manual",
            palette_values = my_palette)

Briefly, the issue with groups exceeding the maximum in the palette is solved by a check which uses default ggplot2 colours if the palette is exceeded.

The issue of more than 9 groups causing errors was fixed by factoring the x positions so that they are in numeric order (i.e. 10 follows 9 rather than 1). However this also required naming the palette so that it can be applied correctly to Cumming plots (as these have missing categories). This choice meant that apply_palette() had to be rewritten. The only alternative would've been to manually colour each bar and violin as is done for the text markers and this felt too hacky to be a solution.