Color Theory

CSI-MTH-190

Schwab

Libraries

library(tidyverse)
library(palmerpenguins) 

Goal: Communication

Remember the goal of making data graphics to to communicate some idea.

In this lecture we’ll talk about different color palettes and when to use them.

We’ll use the penguins data again.

Colors

Big Idea: Color Choices Matter

General Color Palettes:

• Sequential Palettes

• Diverging Palettes

• Qualitative Palettes

R Color Brewer palletts

The two functions

The two functions we’ll use for colors are

scale_*_brewer() scale_*_distiller()

The * matches the aestheic we are trying to color.

scale_fill_brewer()     #<- filling discrete

scale_fill_distiller()  #<- filling continuous

scale_color_brewer()

scale_color_distiller()

Base Graph

Below is a scatter plot of body_mass_g vs bill_depth_mm.

penguins |>
  ggplot()+
  geom_point(aes(x=body_mass_g,bill_depth_mm))

Sequential 1

I’d like to show the flipper length as a third variable. flipper_length_mm is a continuous and numeric variable. We can color by it.

penguins |>
  ggplot()+
  geom_point(aes(
    x = body_mass_g, 
    y = bill_depth_mm,
    color = flipper_length_mm))

Sequential 2

The lighter color should indicate less.

To reverse it I’ll add scale_color_distiller(direction = 1) because we’re working with a continuous variable and we’re coloring.

penguins |>
  ggplot()+
  geom_point(aes(
    x = body_mass_g, 
    y = bill_depth_mm,
    color = flipper_length_mm))+
  scale_color_distiller(direction = 1)

Sequential 3

Using the default color isn’t great. Let’s choose a premade brewer palette.

To reverse it I’ll add scale_color_distiller(direction = 1) because we’re working with a continuous variable and we’re coloring.

penguins |>
  ggplot()+
  geom_point(aes(
    x = body_mass_g, 
    y = bill_depth_mm,
    color = flipper_length_mm))+
  scale_color_distiller(palette = "Greys",  direction = 1)

Sequential

Notes:

• Used for continuous numeric variables

• Darker should be used to show more

Diverging Palettes

If we want to use two base colors, one to show high values and the other to show low. We can use a diverging palette.

Here the brown is long flipper and dark blue is short flipper.

penguins |>
  ggplot()+
  geom_point(aes(
    x = body_mass_g, 
    y = bill_depth_mm,
    color = flipper_length_mm))+
  scale_color_distiller(palette = "BrBG")

Diverging Palettes 2

It often makes more sense to use a diverging palette with the middle being zero.

Here I do a bit of wrangling to add a new column, which captures the difference from the average height.

ave_mass = mean(penguins$body_mass_g,na.rm = TRUE)

penguins |>
  mutate(diff_mass_from_avg =body_mass_g - ave_mass)|>
  ggplot()+
  geom_point(aes(
    x = flipper_length_mm, 
    y = bill_depth_mm,
    color = diff_mass_from_avg))+
  scale_color_distiller(palette = "BrBG")

Diverging Palettes 3

I’m adding a circle around each point so they are easier to see.

ave_mass = mean(penguins$body_mass_g,na.rm = TRUE)

penguins |>
  mutate(diff_mass_from_avg =body_mass_g - ave_mass)|>
  ggplot()+
  geom_point(aes(
    x = flipper_length_mm, 
    y = bill_depth_mm,
    fill = diff_mass_from_avg,
    ),
    shape = 21,   # <- this is a circle that allows outlining
    color = "black",  # <- outline color
    stroke = 0.3,    # <- outline thickness
    size = 3)+
  scale_fill_distiller(palette = "BrBG")

Diverging Palettes 4

Missing Data should be clearly represented with grey or a like color.

library(tidycensus)

#Make Mass Median Income
mass_med_income <- 
  get_acs(
    geography = "county", 
    variables = c(med_household_income = "B19013_001"),
    state = "MA",
    # We'll need to pivot wider, but that doesn't work with simple features. 
    geometry = TRUE
  )|>
  # Add centroids to each region using purrr package
 tidyr::separate(NAME, c("County", "State"), sep = ", ") |> 
 tidyr::separate(County, c("County", "Fluff"), sep = " ") |>
  mutate(med_household_income = estimate) |>
  mutate(med_income_discrete = cut_interval(
    med_household_income, 
    length = 15000, 
    labels = c("60-75", "75-90","90-105", "105-120","120-135")
    ))

mass_med_income$med_income_discrete[8] = NA

ggplot() +
  geom_sf(data = mass_med_income, aes(fill = med_income_discrete)) +
 # geom_text(data = mass_pop, aes(x = lng, y = lat, label = County)) +
  labs(fill = "Median Income\n thousands dollars",
       title = "Massachusetts Med Income 2020 ",
       #subtitle = "Median income in 2020 was $91,842",
       subtitle = "Hampshire county data is missing",
       caption = "deptofnumbers.com")+
  scale_fill_brewer(palette = "PuOr",na.value ="grey" )+
  theme_classic()+
  theme_void()

Qualitative Data

If the variable is categorical use a qualitative palette.

Limit yourself to six categories.

fun_facts <- tribble(
  ~fun_facts,
  "County of Mass Bay Colony",
  "County of Plymouth Colony",
  "Newest",
  "Island",
  "County of Mass Bay Colony",
  "County of Plymouth Colony",
  "Newest",
  "I live here",
  "County of Mass Bay Colony",
  "Newest",
  "County of Plymouth Colony",
  "Boarders Three States",
  "Boarders Three States",
  "Island"
)

ggplot() +
  geom_sf(data = bind_cols(mass_med_income, fun_facts), aes(fill = fun_facts)) +
  labs(fill = "",
       title = "Massachusetts Med Income 2020 ",
       subtitle = "Fun Facts about MA Counties",
       caption = "source: https://en.wikipedia.org/wiki/\nList_of_counties_in_Massachusetts")+
  scale_fill_brewer(palette = "Set2",na.value ="grey" )+
  theme_classic()+
  theme_void()

How do we do this?

Categorical/discrete variable

add scale_*_brewer() to your plot

Numeric/continuous Variable

add scale_*_distiller() to your plot

Common Pitfalls

Too many categories and Rainbows!

This is the default option in R

ggplot() +
  geom_sf(data = bind_cols(mass_med_income, fun_facts), aes(fill = County )) +
 # geom_text(data = mass_pop, aes(x = lng, y = lat, label = County)) +
  labs(fill = "County Names",
       title = "Counties in MA")+
  theme_classic()+
  theme(axis.title.x=element_blank(),
        axis.text.x=element_blank(),
        axis.ticks.x=element_blank(),
        axis.title.y=element_blank(),
        axis.text.y=element_blank(),
        axis.ticks.y=element_blank(),
        axis.line = element_blank()
        )

Confusing fill and color

Don’t confuse fill and color. If aes(fill) then you must use scale_fill_brewer() not scale_color_brewer(), and vice versa.

ggplot() +
  geom_sf(data = mass_med_income, aes(fill = med_income_discrete)) +
  labs(fill = "Median Income\n thousands dollars",
       title = "Tried to color the fill aes()",
       subtitle = "Got the default colors",
       caption = "deptofnumbers.com")+
  scale_color_brewer(palette = "PuOr",na.value ="grey" )+
  theme_classic()+
  theme_void()