Palatable Palettes: Creating and applying bespoke colour schemes

Cara Thompson | NHS-R 2022 Conference | 17th November 2022

Hi there 👋 !

👩 Cara Thompson

👩‍💻 Freelance data consultant specialising in dataviz and “enhanced” reproducible outputs

💙 Helping others maximise the impact of their expertise

🎨 Five tips for creating and applying bespoke colour schemes to your plots

Please suspend all disbelief

The Great Penguin Bake Off

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Make use of colour interpolation

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Make use of colour interpolation

#1 - Use colour purposefully

Make it easy for the readers to remember what is what.

But my graphs aren’t about penguins baking banana bread!

#1 - Use colour purposefully

Make it easy for the readers to remember what is what.

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Make use of colour interpolation

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Make use of colour interpolation

#2 - Let others help you

🤫 - I find picking colours really tricky

Luckily, there are others who can help…

Your department brand guidelines
A photo + something like imagecolorpicker.com to pick out colours

#2 - Let others help you

#2 - Let others help you

#2 - Let others help you

#2 - Let others help you

Your department brand guidelines
A photo + something like imagecolorpicker.com to pick out colours
Take inspiration from other dataviz / art you like

#2 - Let others help you

Your department brand guidelines
A photo + something like imagecolorpicker.com to pick out colours
Take inspiration from other dataviz / art you like
Google images and “[whatever you like] palette”

#2 - Let others help you

Your department brand guidelines
A photo + something like imagecolorpicker.com to pick out colours
Take inspiration from other dataviz / art you like
Google images and “[whatever you like] palette”
Or… start from the colour wheel and read around how best to use it
- Using a tool like paletton.com makes it easier!

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Package up as a default palette

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Package up as a default palette

#3 - Apply colours using a named vector

Once you’ve found your colours, the quick fix…

ggplot(penguins, 
       aes(x = bill_depth_mm,
           y = bill_length_mm,
           colour = species)) +
  geom_point(size = 5) +
  labs(x = "Baking duration",
       y = "Yumminess") +
  theme_nhsr_demo() +
  scale_colour_manual(values = c("#89973d", 
                                 "#e8b92f", 
                                 "#a45e41"))

#3 - Apply colours using a named vector

… might be a dangerous shortcut!

ggplot(penguins %>% 
         # Oh, that should be a factor, 
         # let me fix that for you!
         mutate(species = 
                  factor(species, 
                         levels = c("Chinstrap", 
                                    "Gentoo", 
                                    "Adelie"))), 
       aes(x = bill_depth_mm,
           y = bill_length_mm,
           colour = species)) +
  geom_point(size = 5) + 
  labs(x = "Baking duration",
       y = "Yumminess") +
  theme_nhsr_demo() +
  scale_colour_manual(values = c("#89973d", 
                                 "#e8b92f", 
                                 "#a45e41"))

#3 - Apply colours using a named vector

What is going on?

The values will be matched in order (usually alphabetical) with the limits of the scale, or with breaks if provided. If this is a named vector, then the values will be matched based on the names instead. Data values that don’t match will be given na.value.

#3 - Apply colours using a named vector

Create a named vector

banana_colours <- c("Adelie" = "#89973d",
                    "Chinstrap" = "#e8b92f",
                    "Gentoo" = "#a45e41")

ggplot(penguins,
       aes(x = bill_depth_mm,
           y = bill_length_mm,
           colour = species)) +
  geom_point(size = 5) +
  labs(x = "Baking duration",
       y = "Yumminess") +
  theme_nhsr_demo() +
  scale_colour_manual(values = banana_colours)

#3 - Apply colours using a named vector

Create a named vector

banana_colours <- c("Adelie" = "#89973d",
                    "Chinstrap" = "#e8b92f",
                    "Gentoo" = "#a45e41")

ggplot(penguins %>% 
         # Oh, that should be a factor, 
         # let me fix that for you!
         mutate(species = 
                  factor(species, 
                         levels = c("Chinstrap", 
                                    "Gentoo", 
                                    "Adelie"))), 
       aes(x = bill_depth_mm,
           y = bill_length_mm,
           colour = species)) +
  geom_point(size = 5) +
  labs(x = "Baking duration",
       y = "Yumminess") +
  theme_nhsr_demo() +
  scale_colour_manual(values = banana_colours)

#3 - Apply colours using a named vector

Key advantages

Know the colours are applied to the right data points!
Keep colour-data pairings consistent throughout the project
Package up a default palette
Reuse colours in the text
- ggtext::element_markdown()- Level up your plots

#3 - Apply colours using a named vector

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Package up as a default palette

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Package up as a default palette

#4 - Check for accessibility

📦 {colorblindr} - Claire D. McWhite and Claus O. Wilke

a package to apply simulations of color vision deficiencies to existing ggplot2 figures. It can simulate various classes and severities of color blindness, as well as desaturate plots.

#4 - Check for accessibility

banana_plot

#4 - Check for accessibility

colorblindr::cvd_grid()

#4 - Check for accessibility

Easiest way to do this? Build a palette that goes from a lighter colour to a darker colour.

A few helpful resources:

Find colours - mycolor.space
Test palette - Chroma.js Color Palette Helper
Test text contrast - webaim.org/resources/contrastchecker

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Make use of colour interpolation

#1 - Use colour purposefully

#2 - Let others help you

#3 - Apply colours using a named vector

#4 - Check for accessibility

#5 - Make use of colour interpolation

What if I don’t know how many colours I need?

Build more than you’re likely to need and allow R to pick colours to maximise distance
- ⚠️ It just picks e.g. #1, #5 and #9 in the order of your vector
Pick 2 or 3 anchor colours and let R interpolate
- colorRampPalette()

#5 - Make use of colour interpolation

my_anchor_colours <- c("green", "blue", "purple")

palmerpenguins::penguins %>%
  mutate(island_fm = paste0(island, sex)) %>%
  ggplot() + 
  geom_point(aes(x = bill_length_mm,
                 y = bill_depth_mm,
                 colour = island_fm,
                 size = body_mass_g),
             show.legend = FALSE) +
  theme_nhsr_demo() +
  scale_color_manual(
    values = colorRampPalette(
      my_anchor_colours)(9))

#5 - Make use of colour interpolation

imagecolorpicker.com + Chroma.js Color Palette Helper

#5 - Make use of colour interpolation

my_anchor_colours <- c("#4f3c78", "#d3970a", "#c0979c")

palmerpenguins::penguins %>%
  ggplot() + 
  geom_point(aes(x = bill_length_mm,
                 y = bill_depth_mm,
                 colour = island,
                 size = body_mass_g),
             show.legend = FALSE) +
  theme_nhsr_demo() +
  scale_color_manual(
    values = colorRampPalette(
      my_anchor_colours)(3))

#5 - Make use of colour interpolation

my_anchor_colours <- c("#4f3c78", "#d3970a", "#c0979c")

palmerpenguins::penguins %>%
  mutate(island_fm = paste0(island, sex)) %>%
  ggplot() + 
  geom_point(aes(x = bill_length_mm,
                 y = bill_depth_mm,
                 colour = island_fm,
                 size = body_mass_g),
             show.legend = FALSE) +
  theme_nhsr_demo() +
  scale_color_manual(
    values = colorRampPalette(
      my_anchor_colours)(9))

#5 - Make use of colour interpolation

my_anchor_colours <- c("#375248", "#d3970a", "#7691b1")

palmerpenguins::penguins %>%
  mutate(island_fm = paste0(island, sex)) %>%
  ggplot() + 
  geom_point(aes(x = bill_length_mm,
                 y = bill_depth_mm,
                 colour = island_fm,
                 size = body_mass_g),
             show.legend = FALSE) +
  theme_nhsr_demo() +
  scale_color_manual(
    values = colorRampPalette(
      my_anchor_colours)(9))

#5 - Make use of colour interpolation

There are only so many colours you can use in a plot!

Could other approaches work (facets?)

Try using different shapes (you can use letters / emoji!)

There’s so much more we could talk about!

Packaging up your palette for easy reuse carartemplates::carar_colours()

Creating your own scale_nhsdemo_colour/fill() functions
You can even have several palettes! scale_nhsdemo_colour("Birmingham")?

Basing your “dark text” colour on the central / most important colour in your palette

… but we can talk about this some other time!