6 Plot Cosmetics

6.1 Introduction

Beyond mapping data to aesthetics, ggplot2 provides extensive control over the visual appearance of your plots. This chapter covers the “cosmetic” aspects: coordinate systems, position adjustments, themes, labels, legends, and axis formatting.

Note: This chapter introduces many concepts, but not all are equally important. Some (like zooming with coord_cartesian() and using built-in themes) are frequently useful, while others (like polar coordinates or customising individual theme elements) are more specialised. See the summary at the end of this chapter for a priority guide.

We continue to follow the grammar of graphics: first we create layers using geoms (Chapter 3), then we add scale layers to control aesthetics (Chapters 4 and 5), and now we add coordinate systems and themes to control the overall appearance.

6.2 Incremental workflow

A powerful feature of ggplot2 is that you can build plots incrementally by saving them as objects and adding layers or modifications later. This is covered in detail in Practical 5, but the key ideas are:

1. Save plots as objects: p <- ggplot(...) + geom_point() + ...
2. Add layers later: p + theme_minimal() or p + scale_colour_viridis_d()
3. Use last_plot() interactively: In the console, last_plot() returns the most recently created plot, allowing quick refinements

This approach is particularly useful when experimenting with different colour schemes, themes, or labels without rewriting the entire plot code.

6.3 Coordinate systems

The coordinate system determines how x and y positions are interpreted and displayed.

6.3.1 Cartesian coordinates (default)

The default coord_cartesian() uses standard Cartesian coordinates. Its main use is to zoom into a plot without removing data.

Recall from Section 5.5 that setting axis limits with scale_*_continuous(limits = ) removes data outside the specified range. This can be problematic if you later add fitted lines or smoothers, which would be computed from the truncated data. In contrast, coord_cartesian() zooms visually while preserving all data for calculations:

ggplot(mpg, aes(x = displ, y = hwy)) +
  geom_point() +
  coord_cartesian(xlim = c(2, 6), ylim = c(15, 40)) +
  labs(x = "Engine Displacement (L)", y = "Highway MPG")

Figure 6.1: Zooming with coord_cartesian() preserves all data.

Compare this to the example in Section 5.5, which used the same limits with scale_*_continuous(). The visual result is similar, but coord_cartesian() is safer when you need all data points for statistical layers. See also Section 7.4 in Chapter 7 for a side-by-side comparison of all three approaches, including dplyr::filter().

6.3.2 Fixed aspect ratio

Use coord_fixed() to enforce a specific aspect ratio:

ggplot(mpg, aes(x = cty, y = hwy)) +
  geom_point() +
  coord_fixed(ratio = 1) +
  labs(x = "City MPG", y = "Highway MPG")

Figure 6.2: Fixed 1:1 aspect ratio.

6.3.3 Flipped coordinates

coord_flip() swaps x and y axes. This is useful for horizontal bar charts:

ggplot(mpg, aes(x = class)) +
  geom_bar() +
  coord_flip() +
  labs(x = "Vehicle Class", y = "Count")

Figure 6.3: Horizontal bar chart using coord_flip().

Note: After flipping, the x argument in labs() refers to what appears as the vertical axis, and y refers to the horizontal axis. This is correct but can be confusing — you label based on the original mapping, not the visual appearance.

6.3.4 Polar coordinates

coord_polar() transforms Cartesian coordinates to polar coordinates. A pie chart is essentially a stacked bar chart wrapped into a circle:

ggplot(mpg, aes(x = "", fill = drv)) +
  geom_bar(width = 1) +
  coord_polar(theta = "y") +
  labs(fill = "Drive Type") +
  theme_void()

Figure 6.4: Pie chart using coord_polar() — a bar chart wrapped into a circle.

Here, geom_bar() creates a stacked bar (counting observations), and coord_polar(theta = "y") wraps it into a circle where the angle represents the count.

Note: Pie charts are generally discouraged because comparing angles is harder than comparing lengths. A simple bar chart is almost always easier to read.

6.4 Position adjustments

Position adjustments control how overlapping geoms are handled. We have already seen two common position adjustments in Chapter 3:

• Jittering with geom_jitter() to reveal overlapping points (Section 3.7.2)
• Dodging with position = "dodge" to place bars side by side

Here we provide more detail and additional options.

6.4.1 Jittering (for overplotting)

When points overlap, adding small random noise reveals the underlying distribution. Use geom_jitter() or position_jitter():

ggplot(mpg, aes(x = drv, y = hwy)) +
  geom_jitter(width = 0.2, height = 0) +
  labs(x = "Drive Type", y = "Highway MPG")

Figure 6.5: Jittered points to reveal overplotting.

The width and height arguments control the amount of jitter. Setting height = 0 prevents vertical jitter, which is appropriate when the \(y\)-axis represents measured values.

6.4.2 Dodging (side by side)

Dodging places bars (or other geoms) side by side rather than stacked. We have already seen this in Section 3.8.1 for grouped bar charts:

ggplot(mpg, aes(x = drv, fill = factor(cyl))) +
  geom_bar(position = "dodge") +
  labs(x = "Drive Type", y = "Count", fill = "Cylinders")

Figure 6.6: Dodged bars (side by side).

Use position_dodge2(preserve = "single") for bars of equal width when some groups are missing:

ggplot(mpg, aes(x = drv, fill = factor(cyl))) +
  geom_bar(position = position_dodge2(preserve = "single")) +
  labs(x = "Drive Type", y = "Count", fill = "Cylinders")

Figure 6.7: Dodged bars with equal width.

6.4.3 Stacking (default for bars)

The default position for bars is stacking:

ggplot(mpg, aes(x = drv, fill = factor(cyl))) +
  geom_bar(position = "stack") +
  labs(x = "Drive Type", y = "Count", fill = "Cylinders")

Figure 6.8: Stacked bars (default).

6.4.4 Filling (proportions)

position = "fill" scales bars to equal height, showing proportions:

ggplot(mpg, aes(x = drv, fill = factor(cyl))) +
  geom_bar(position = "fill") +
  labs(x = "Drive Type", y = "Proportion", fill = "Cylinders")

Figure 6.9: Proportional stacked bars.

6.5 Themes

Themes control the non-data elements of a plot: background, grid lines, fonts, etc.

6.5.1 Built-in themes

ggplot2 provides several complete themes. Each theme function accepts a base_size argument to control the overall font size (default is 11):

p <- ggplot(mpg, aes(x = displ, y = hwy)) +
  geom_point() +
  labs(x = "Engine Displacement (L)", y = "Highway MPG")

p + theme_grey(base_size = 14)  # default theme, larger text

Figure 6.10: Default theme_grey().

p + theme_bw(base_size = 14)

Figure 6.11: theme_bw() - black and white.

p + theme_minimal(base_size = 14)

Figure 6.12: theme_minimal() - minimal.

p + theme_classic(base_size = 14)

Figure 6.13: theme_classic() - classic.

Increasing base_size is an easy way to make plots more readable, especially for presentations or publications.

6.5.2 Setting a global theme

Use theme_set() to change the default theme for all subsequent plots. You can also set a larger base_size globally:

theme_set(theme_minimal(base_size = 14))

This is particularly useful in R Markdown documents where you want consistent styling across all plots.

6.5.3 Customising individual elements

The theme() function has many arguments for customising specific elements of a plot: titles, axis labels, tick labels, grid lines, backgrounds, legends, and more. While these options exist, I recommend focusing on overall legibility (through base_size and built-in themes) rather than spending too much time on fine-grained customisation. The built-in themes are well-designed and sufficient for most purposes. The one use of theme() that I would highlight is for the legend position - see Section 6.7.

6.6 Labels

We have already used labs() throughout this module for axis labels and legend titles. Here we show that labs() can also add titles, subtitles, and captions:

ggplot(mpg, aes(x = displ, y = hwy, colour = factor(cyl))) +
  geom_point(size = 3) +
  labs(
    x = "Engine Displacement (L)",
    y = "Highway MPG",
    colour = "Cylinders",
    title = "Fuel Efficiency by Engine Size",
    subtitle = "Data from mpg dataset",
    caption = "Source: EPA fuel economy data"
  )

Figure 6.14: Using labs() for titles, subtitles, and captions.

Note how labs(colour = "Cylinders") fixes the ugly legend title that would otherwise appear as “factor(cyl)”. Whenever you map an aesthetic to a transformed variable (like factor(cyl) or log(value)), use labs() to provide a clean title.

6.7 Legends

Legends are generated automatically when you map variables to aesthetics. You can control their position using theme(legend.position = ).

6.7.1 Legend position

ggplot(mpg, aes(x = displ, y = hwy, colour = drv)) +
  geom_point(size = 3) +
  labs(x = "Engine Displacement (L)", y = "Highway MPG", colour = "Drive") +
  theme(legend.position = "bottom")

Figure 6.15: Legend positioned at bottom.

Options for legend.position:

• "top", "bottom", "left", "right": place legend outside the plot area
• "none": remove the legend entirely
• "inside": place legend inside the plot area (see below)

6.7.2 Legend inside the plot

To place a legend inside the plot area, use legend.position = "inside" along with legend.position.inside to specify the coordinates:

ggplot(mpg, aes(x = displ, y = hwy, colour = drv)) +
  geom_point(size = 3) +
  labs(x = "Engine Displacement (L)", y = "Highway MPG", colour = "Drive") +
  theme(
    legend.position = "inside",
    legend.position.inside = c(0.9, 0.8),
    legend.background = element_rect(fill = "white", colour = "grey80")
  )

Figure 6.16: Legend inside the plot area.

The coordinates c(0.9, 0.8) specify the position as fractions of the plot area (0 = left/bottom, 1 = right/top).

6.8 Axis formatting

6.8.1 Avoiding scientific notation

Large numbers are sometimes displayed in scientific notation (e.g., 1e+06 instead of 1,000,000). To use standard notation with comma separators, use scales::label_comma():

# Create data with large values
big_data <- data.frame(
  x = c(100000, 500000, 1000000, 2000000),
  y = c(10, 25, 40, 55)
)

ggplot(big_data, aes(x = x, y = y)) +
  geom_point(size = 3) +
  scale_x_continuous(labels = scales::label_comma()) +
  labs(x = "Population", y = "Value")

Figure 6.17: Avoiding scientific notation on axes.

Other useful label functions from the scales package:

Function	Effect
label_comma()	Adds comma separators: 1,000,000
label_dollar()	Adds currency symbol: $1,000
label_percent()	Converts to percentage: 50%
label_scientific()	Forces scientific notation

6.8.2 Controlling tick marks

Use scale_*_continuous() to control where tick marks appear (see also Section 5.5 for axis transformations):

ggplot(mpg, aes(x = displ, y = hwy)) +
  geom_point() +
  scale_x_continuous(breaks = seq(2, 7, by = 1)) +
  scale_y_continuous(breaks = seq(15, 45, by = 5)) +
  labs(x = "Engine Displacement (L)", y = "Highway MPG")

Figure 6.18: Custom tick mark positions.

6.9 Summary

The following table summarises the topics covered in this chapter with their priority for this module:

Topic	Section	Priority
Cartesian coordinates (zooming)	6.3.1	High
Fixed aspect ratio	6.3.2	Medium
Flipped coordinates	6.3.3	Medium
Polar coordinates	6.3.4	Low
Position adjustments	6.4.2	Medium (already covered)
Built-in themes	6.5.1	Medium
Setting a global theme	6.5.2	Medium
Customising individual elements	6.5.3	Low
Labels	6.6	Medium (already covered)
Legend position	6.7.1	Medium
Legend inside the plot	6.7.2	Low
Avoiding scientific notation	6.8	Low
Controlling tick marks	6.8	Low (for now)

Key takeaways:

• Use coord_cartesian() to zoom without removing data
• Use theme_*() functions with base_size for easy, consistent styling
• Use labs() to fix legend titles and add titles/captions
• Position adjustments ("dodge", "jitter") were introduced in Chapter 3