3.3 ggplot2: Facets, Themes, and Saving

1. Facets

Facets make a panel for each level of a categorical variable. That is often clearer than twenty colors on one panel.

facet_wrap(~ class) wraps one variable into a grid:

ggplot(data = mpg, mapping = aes(x = displ, y = hwy)) +
  geom_point() +
  facet_wrap(~ class)

facet_grid(rows ~ cols) is a matrix of two categoricals:

ggplot(data = mpg, mapping = aes(x = displ, y = hwy)) +
  geom_point(alpha = 0.3) +
  facet_grid(drv ~ class)

A . means “no faceting on that side”:

ggplot(data = mpg, mapping = aes(x = displ, y = hwy)) +
  geom_point() +
  facet_grid(class ~ .)

ggplot(data = mpg, mapping = aes(x = displ, y = hwy)) +
  geom_point() +
  facet_grid(. ~ drv)

2. A fixed color is not a mapping

To make every point blue, set color outside aes():

ggplot(data = mpg, mapping = aes(x = displ, y = hwy)) +
  geom_point(color = "blue")

The name is a character, so it is in quotes. Inside aes(), ggplot2 treats "blue" as a one-level categorical variable and draws a legend for it. That is a common mistake:

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = "blue")) +
  geom_point()

The same rule applies to histograms. fill is the interior; color is the outline:

ggplot(data = mpg, mapping = aes(x = hwy)) +
  geom_histogram(bins = 25, fill = "white", color = "black")

3. Themes

A theme sets the background, grid, and fonts in one step. theme_bw() is a usual choice for a document. theme_classic() is quieter.

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = class)) +
  geom_point() +
  theme_bw()

ggthemes adds more, including theme_economist() and theme_fivethirtyeight(). Load the package first.

4. Scales and colorblind palettes

A scale maps data values to visual values. You can set the ends of a continuous color scale, or log both axes:

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = cty)) +
  geom_point() +
  scale_color_continuous(low = "black", high = "white")

ggplot(data = diamonds, mapping = aes(x = carat, y = price)) +
  geom_point(alpha = 0.01) +
  scale_x_log10() +
  scale_y_log10()

For a plot you will share, use a colorblind-safe palette. ggthemes::scale_color_colorblind() and scale_fill_colorblind() are one option. viridis is another: colorful, perceptually even, and still readable in grayscale. Use scale_color_viridis_d() / scale_fill_viridis_d() for discrete groups and _c for continuous.

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = class)) +
  geom_point() +
  theme_bw() +
  scale_color_colorblind()

ggplot(data = mpg, mapping = aes(x = hwy, fill = drv)) +
  geom_histogram(position = "dodge", bins = 15) +
  theme_bw() +
  scale_fill_viridis_d()

5. Labels

xlab(), ylab(), and ggtitle() set the axis titles and the main title. The legend title is an argument on the scale (name = ...).

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = class)) +
  geom_point() +
  theme_bw() +
  scale_color_discrete(name = "Car Class") +
  xlab("Displacement") +
  ylab("Highway") +
  ggtitle("Displacement vs Highway")

6. Saving a plot

A ggplot is an object. Assign it, print it, then export.

pl <- ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = class)) +
  geom_point() +
  theme_bw()

pl

In RStudio, Export in the Plots pane writes a file. In a script or Quarto document, ggsave() does the same. It overwrites without asking.

ggsave("hwy-vs-displ.pdf", plot = pl, width = 6, height = 4)

If you omit plot, ggsave() uses the last plot you printed. Put the file in a folder you already have, or create one first.

7. This week, in one place

By the end of Week 3 you should be able to:

• build a plot with ggplot(), aes(), and a geom_*()
• choose a geom from the variables: scatter (two quantitative), histogram or density (one quantitative), boxplot (one categorical and one quantitative)
• map a third variable with color, fill, or facets
• add a smooth, fix overplotting, change a theme and a scale
• label the plot and save it with ggsave()

Functions from this week include ggplot(), aes(), geom_point(), geom_smooth(), geom_jitter(), geom_hex(), geom_histogram(), geom_density(), geom_boxplot(), facet_wrap(), facet_grid(), theme_bw(), scale_y_log10(), xlab(), ylab(), ggtitle(), and ggsave().

Chapter 3 of R for Data Science and the ggplot2 cheat sheet are the usual references.

8. Practice

1. Boxplots of diamond price vs cut, log-y, faceted by color.

2. The same plot: orange boxes, theme_bw().

3. Save that plot with ggsave().