Lecture 4

ggplot2: Scales, Guides, and Themes (How to Control What We See)

Byeong-Hak Choe

bchoe@geneseo.edu

SUNY Geneseo

February 28, 2026

Scales: scale_*_*()

What is a scale_*_*()?

A scale control aesthetics mappings:

• x/y position: scale_x_*(), scale_y_*()

• color/fill: scale_color_*(), scale_fill_*()

• size/shape/alpha: scale_size_*(), scale_shape_*(), scale_alpha_*()

• Each deals with one combination of mapping and scale. Too many to memorize (131 distinct scale_*_*() functions)!

  • They are named according to a consistent logic

• https://ggplot2tor.com/apps provides a complete guide to scales and themes, as well as aesthetics.

Scale Kinds (Choose Based on Our Variable Types)

• Continuous: numeric with many possible values
  → *_continuous(), *_log10(), *_sqrt(), *_date(), *_datetime()

• Discrete: categories / factors
  → *_discrete(), *_manual()

• Rule of thumb:

  • num/int → continuous
  • factor/chr → discrete

scale_x_continuous(): breaks & labels

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(x = gdpPercap, 
             y = lifeExp,
             color = continent)) +
  geom_point(alpha = 0.8,
             size = 3) +
  geom_smooth(se = F) +
  scale_x_continuous(
    breaks = c(1000, 10000, 50000),
    labels = scales::dollar
  ) +
  labs(
    x = "GDP per capita",
    y = "Life expectancy",
    color = "Continent")

What this controls:

• Where ticks appear (breaks)
• How ticks display text (labels)

Common scale_*() Arguments You’ll Use a Lot

For most scales, these are the “big four”:

• name = legend/axis title (often we can do this in labs() too)
• limits = c(min, max) what values are shown
• breaks = where tick marks / legend entries appear
• labels = how ticks / legend entries display as text

Limits vs. Zooming: limits vs coord_*

Two different ideas:

A) Scale Limits (can drop data)

... + scale_x_continuous(limits = c(0, 20000))

B) Coordinate Zoom (keeps data, just views a window)

... + coord_cartesian(xlim = c(0, 20000))

• With limits =, ggplot fits the smooth using only data (\(\leq\) 20,000) (outliers removed).
• With coord_cartesian(), ggplot fits the smooth using all data, then just shows the part of the line within the window.

Color Scales: Discrete vs. Continuous

Discrete Color (Categorical):

... + scale_color_viridis_d()

Continuous Color (Numeric Gradient):

... + scale_color_viridis_c()

Common mistake:

• Numeric variable that is really a category (e.g., year)
  • Fix with aes(color = factor(year)) or convert to factor

Manual Scales: When We Must Choose Specific Values

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(x = gdpPercap, 
             y = lifeExp, 
             color = continent)) +
  geom_point(size = 3) +
  scale_color_manual(
    values = c(
      "Africa" = "#1b9e77",
      "Americas" = "#d95f02",
      "Asia" = "#7570b3",
      "Europe" = "#e7298a",
      "Oceania" = "#66a61e"
    )
  )

• A ton of hex codes for colors are available here:
  • https://www.color-hex.com

Multiple Scales: One Plot Can Have Many

A single plot can have:

• x scale
• y scale
• color scale
• fill scale
• shape scale
• size scale
• …

Each mapped aesthetic typically has one scale.

🎨 RColorBrewer

Use Color to Your Advantage

• Choose a color palette based on the type of data you are plotting:
  • Unordered categorical variable (e.g., gender, country) → Distinct colors that won’t be easily confused
  • Ordered categorical variable (e.g., Level of Education) → Graded color scheme running from less to more or earlier to later

RColorBrewer

• RColorBrewer provides a wide range of named color palettes.
• Access colors using scale_color_brewer() or scale_fill_brewer() with the palette parameter.

Sequential

• Sequential palettes are suited to ordered data that progress from low to high.

Diverging

• Diverging palettes put equal emphasis on mid-range critical values and extremes at both ends of the data range.

Qualitative

• Qualitative palettes do not imply magnitude differences between legend classes.

• Qualitative schemes are best suited to representing unordered categorical data.

Display All Color Palettes

display.brewer.all()

scale_color_brewer(palette = ...)

p <- ggplot(data = socviz::organdata,
            mapping = 
              aes(x = roads, 
                  y = donors, 
                  color = world))
p + geom_point(size = 2) + 
  scale_color_brewer(
    palette = "Set2") +
  theme(legend.position = "top")

• Use named palettes with scale_color_brewer(palette = ...).

Pastel2

p + geom_point(size = 3) + 
  scale_color_brewer(
    palette = "Pastel2") +
  theme(legend.position = "top")

Dark2

p + geom_point(size = 3) + 
  scale_color_brewer(
    palette = "Dark2") +
  theme(legend.position = "top")

Color-Blindness Friendly Palettes

• RColorBrewer flags palettes that are color-blind friendly in the colorblind column.
• Prefer these palettes (e.g., "Set2", "Dark2") when designing for broad audiences.

Getting Hex Color Codes from a Palette

# Get 5 hex codes from the "Set2" palette
brewer.pal(n = 5, name = "Set2")

• Use brewer.pal(n, name) to extract hex color codes from any named palette:
  • n — number of colors to extract (must be between 3 and the palette’s maximum)
  • name — name of the palette (e.g., "Set2", "Blues", "RdBu")

brewer.pal.info  # check the 'maxcolors' column

• brewer.pal.info provides the maximum number of colors available per palette

scale_*_manual() with brewer.pal()

my_colors <- 
  brewer.pal(n = 4, 
             name = "Dark2")

p + geom_point(size = 3) +
  scale_color_manual(
    values = my_colors) +
  theme(legend.position = "top")

• You can then pass hex codes directly into scale_color_manual() or scale_fill_manual().

Guides: guides()

What is a Guide?

A guide is the display of a scale:

• discrete color → legend
• continuous color → colorbar
• size → legend
• shape → legend

Guides answer:

• “What does this color mean?”
• “Which categories exist?”
• “What value corresponds to this gradient?”

Remove a Legend (most common use)

If we don’t want a legend for an aesthetic:

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = continent)) +
  geom_point(size = 3) +
  guides(color = "none")

Equivalent shortcut:

... + theme(legend.position = "none")

Difference:

• guides(color = "none") removes only that guide
• theme(legend.position="none") removes all legends

Choose the Guide Type Explicitly

Sometimes ggplot guesses well; sometimes we want control:

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = pop)) +
  geom_point(alpha = 0.8) +
  scale_color_viridis_c() +
  guides(
    color = 
      guide_colorbar(
        barheight = unit(4, "cm")))

Legend Layout: Rows/Columns

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, color = continent)) +
  geom_point(size = 3) +
  guides(color = guide_legend(nrow = 1, byrow = TRUE))

Useful when moving legend to top/bottom.

Reorder Multiple Legends (order =)

When we map multiple aesthetics, we can control the order:

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = continent, 
             size = pop)) +
  geom_point(alpha = 0.8) +
  guides(
    color = guide_legend(order = 1),
    size  = guide_legend(order = 2)
  )

Override the Legend Appearance (override.aes)

Sometimes the plot uses alpha/size that makes legend hard to read.

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = continent)) +
  geom_point(alpha = 0.2, 
             size = 3) +
  guides(
    color = 
      guide_legend(
        override.aes = list(alpha = 1)))

When Do We Need guides()?

Use guides() when we want to:

• remove a specific legend (not all)
• change legend layout (nrow/ncol, direction)
• force legend vs colorbar
• reorder multiple legends
• override how keys look

If we only want to rename legends, start with labs().

Themes: theme()

What is a Theme?

A theme controls everything that is not our data:

• backgrounds, gridlines
• fonts, text size, text alignment
• spacing/margins
• legend position and styling
• facet strip styling

Themes do not change the data mapping.

Theme “Starter Packs”: theme_*()

Try different base themes quickly:

p <- gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = continent)) +
  geom_point(size = 3) +
  scale_x_log10()

p + theme_minimal()
p + theme_classic()
p + theme_light()
p + hrbrthemes::theme_ipsum()
p + ggthemes::theme_fivethirtyeight()

The Most-Used Theme Edits

Legend Position

... + theme(legend.position = "top")

Remove Minor Gridlines

... + theme(panel.grid.minor = element_blank())

Emphasize Axis Titles

Theme System

Theme Elements: “What Can I Edit?”

The theme() function has 94 possible arguments!

Common ones:

• plot.title, plot.subtitle, plot.caption
• axis.title.x, axis.title.y, axis.text.x, axis.text.y
• panel.grid.major, panel.grid.minor, panel.background
• legend.position, legend.title, legend.text, legend.key
• strip.text, strip.background (facets)
• plot.margin

The only way to learn how to use theme() is to use it and tinker with it.

Example: a Clean “Presentation-Ready” Theme

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = continent)) +
  geom_point(size = 3, alpha = 0.9) +
  scale_x_log10(labels = scales::dollar) +
  labs(
    title = "Life expectancy vs GDP per capita (2007)",
    subtitle = "Log x-axis; color = continent",
    x = "GDP per capita (log scale)",
    y = "Life expectancy",
    color = "Continent",
    caption = "Source: gapminder"
  ) +
  theme(
    legend.position = "top",
    plot.title = element_text(face = "bold"),
    panel.grid.minor = element_blank()
  )

Putting It Together

One Plot, All Three Controls

gapminder |>
  filter(year == 2007) |>
  ggplot(aes(gdpPercap, lifeExp, 
             color = continent, 
             size = pop)) +
  geom_point(alpha = 0.7) +

  # SCALES: transform, breaks, labels
  scale_x_log10(
    breaks = c(500, 2000, 10000, 50000),
    labels = scales::dollar
  ) +
  scale_size_continuous(labels = scales::label_number(scale_cut = scales::cut_si(""))) +

  # GUIDES: legend order + layout
  guides(
    color = guide_legend(order = 1, nrow = 1),
    size   = guide_legend(order = 2)
  ) +

  # THEME: layout + typography
  theme(
    legend.position = "top",
    plot.title = element_text(face = "bold"),
    panel.grid.minor = element_blank()
  ) +

  # labels are not scales/guides/themes, but they coordinate everything
  labs(
    title = "Scales + Guides + Themes in one figure",
    x = "GDP per capita (log scale)",
    y = "Life expectancy",
    color = "Continent",
    size = "Population"
  )

Summary

What to Reach for First ✅

• Data → visual mapping: aes() + scale_*_*()
• Explaining mappings: labs() + guides()
• Polish / layout: theme() (or theme_*())

One-Sentence Definitions

• Scale: converts data values into aesthetic values.
• Guide: shows the scale to the reader (legend/colorbar).
• Theme: controls the look of the plot frame.