Lesson 4 of 7 · Course overview

Data Visualization in R

Plotting in R is excellent. You’ve already seen plot() — it works, but the real toolkit is ggplot2. It’s based on a “grammar of graphics” that, once it clicks, lets you describe almost any plot in a handful of lines.

This lesson covers the ggplot2 grammar, the most common geoms, faceting, themes, and how to save plots.

Setup

library(ggplot2)
library(dplyr)

cars <- mtcars |>
  tibble::rownames_to_column("model") |>
  tibble::as_tibble() |>
  mutate(
    cyl = factor(cyl),
    transmission = if_else(am == 1, "manual", "automatic")
  )

The ggplot2 grammar

Every ggplot2 plot has three required pieces:

Data — the data frame you’re plotting.
Aesthetics (aes()) — which columns map to which visual properties (x, y, color, size, shape).
Geoms — the geometric shapes to draw (points, bars, lines, boxes, …).

The pattern:

ggplot(data, aes(x = ..., y = ..., color = ...)) +
  geom_*()

You add layers with +. Yes, + (not the pipe). Inside a ggplot() call you +, between data steps you |>. It’s a mild gotcha and you’ll get it wrong a lot at first.

A first plot

ggplot(cars, aes(x = wt, y = mpg)) +
  geom_point()

That’s the whole plot: data (cars), aesthetics (wt on x, mpg on y), one geom (geom_point for a scatterplot).

Now let’s add color, size, and labels:

ggplot(cars, aes(x = wt, y = mpg, color = cyl, size = hp)) +
  geom_point(alpha = 0.8) +
  labs(
    title = "Heavier, more powerful cars get worse mileage",
    subtitle = "Each point is a 1974 car model",
    x = "Weight (1000 lbs)",
    y = "Miles per gallon",
    color = "Cylinders",
    size = "Horsepower"
  )

Notice three things:

color = cyl and size = hp are inside aes() because they map to data columns.
alpha = 0.8 is outside aes() because it’s a fixed value (every point gets the same alpha), not a mapping.
labs() controls every text label on the plot at once.

The most useful geoms

`geom_point()` — scatterplots

Shown above. Add geom_smooth() to overlay a regression line:

ggplot(cars, aes(x = wt, y = mpg)) +
  geom_point() +
  geom_smooth(method = "lm", se = TRUE)

method = "lm" is a straight line; the default "loess" is a smooth curve.

`geom_line()` — connect points by some ordering

year_data <- tibble::tibble(
  year = 2015:2024,
  sales = c(100, 108, 121, 119, 130, 145, 160, 158, 172, 185)
)

ggplot(year_data, aes(x = year, y = sales)) +
  geom_line(linewidth = 1.1, color = "steelblue") +
  geom_point(size = 2.5, color = "steelblue") +
  labs(title = "Sales over time", x = NULL, y = "Sales (units)")

`geom_bar()` and `geom_col()` — bar charts

geom_bar() counts rows itself. geom_col() uses the y-value you give it.

ggplot(cars, aes(x = cyl)) +
  geom_bar(fill = "steelblue") +
  labs(title = "How many of each cylinder count?", x = "Cylinders", y = "Count")

cars |>
  group_by(cyl) |>
  summarise(mean_mpg = mean(mpg), .groups = "drop") |>
  ggplot(aes(x = cyl, y = mean_mpg, fill = cyl)) +
  geom_col() +
  geom_text(aes(label = round(mean_mpg, 1)), vjust = -0.4) +
  labs(title = "Average MPG by cylinder count", x = "Cylinders", y = "Mean MPG") +
  theme(legend.position = "none")

Notice we pipe a dplyr summary directly into ggplot(). This is the standard tidyverse workflow.

`geom_histogram()` and `geom_density()` — distributions

ggplot(cars, aes(x = mpg)) +
  geom_histogram(bins = 12, fill = "steelblue", color = "white") +
  labs(title = "Distribution of MPG across all cars", x = "MPG", y = "Count")

ggplot(cars, aes(x = mpg, fill = transmission)) +
  geom_density(alpha = 0.5) +
  labs(title = "MPG by transmission type", x = "MPG", y = "Density")

`geom_boxplot()` — distributions by group

ggplot(cars, aes(x = cyl, y = mpg, fill = cyl)) +
  geom_boxplot() +
  labs(title = "MPG by cylinder count", x = "Cylinders", y = "MPG") +
  theme(legend.position = "none")

Faceting — small multiples

facet_wrap() and facet_grid() let you split a plot into a grid of subplots, one per category. Often more informative than mapping a category to color.

ggplot(cars, aes(x = wt, y = mpg)) +
  geom_point() +
  geom_smooth(method = "lm", se = FALSE) +
  facet_wrap(~ cyl, labeller = label_both) +
  labs(title = "Weight vs. MPG by cylinder count")

facet_grid() does a 2D grid of categories:

ggplot(cars, aes(x = wt, y = mpg)) +
  geom_point() +
  facet_grid(transmission ~ cyl, labeller = label_both) +
  labs(title = "Weight vs. MPG by transmission and cylinder count")

Scales — controlling axes and color

Scales control how data values map to visual properties. Common ones:

ggplot(cars, aes(x = wt, y = mpg, color = hp)) +
  geom_point(size = 3) +
  scale_color_viridis_c(option = "plasma") +
  scale_x_continuous(limits = c(1, 6), breaks = 1:6) +
  scale_y_continuous(name = "Miles per gallon") +
  labs(title = "Weight vs. MPG, colored by horsepower")

A few worth knowing:

scale_*_continuous() — for numeric axes; sets limits, breaks, labels.
scale_*_log10() — log axis (useful when data spans orders of magnitude).
scale_color_viridis_*() / scale_fill_viridis_*() — perceptually uniform, color-blind-friendly palettes. Use these.
scale_color_manual() — assign specific colors to specific levels.

Themes — controlling the non-data ink

Themes change how the plot looks — gridlines, fonts, backgrounds. ggplot2 ships with several:

p <- ggplot(cars, aes(x = wt, y = mpg, color = cyl)) +
  geom_point(size = 3) +
  labs(title = "Weight vs. MPG", x = "Weight", y = "MPG")

p + theme_minimal()

p + theme_classic()

You can also tweak individual theme elements:

p +
  theme_minimal(base_size = 13) +
  theme(
    plot.title = element_text(face = "bold", size = 16),
    legend.position = "bottom",
    panel.grid.minor = element_blank()
  )

Setting a theme for the whole session:

theme_set(theme_minimal(base_size = 13))

Saving plots

ggsave() saves the most recent plot — or one you pass to it — as a file:

my_plot <- ggplot(cars, aes(x = wt, y = mpg)) + geom_point()

ggsave("my_plot.png", my_plot, width = 6, height = 4, dpi = 300)
ggsave("my_plot.pdf", my_plot, width = 6, height = 4)

ggsave() infers the format from the file extension. Width and height are in inches by default.

Putting it together

Let’s build a single, polished figure that summarises something interesting from the data: how MPG depends on weight, broken out by cylinder count, with a fitted line for each group.

ggplot(cars, aes(x = wt, y = mpg, color = cyl)) +
  geom_point(size = 3, alpha = 0.85) +
  geom_smooth(method = "lm", se = FALSE, linewidth = 0.9) +
  scale_color_viridis_d(option = "viridis", end = 0.85) +
  labs(
    title = "Heavier cars get worse mileage — but cylinder count matters too",
    subtitle = "Each point is a 1974 car model from the mtcars dataset",
    x = "Weight (1000 lbs)",
    y = "Miles per gallon",
    color = "Cylinders",
    caption = "Data: Henderson & Velleman (1981). Building multiple regression models interactively."
  ) +
  theme_minimal(base_size = 13) +
  theme(
    plot.title = element_text(face = "bold"),
    plot.subtitle = element_text(color = "grey40"),
    plot.caption = element_text(color = "grey50", hjust = 0),
    legend.position = "bottom",
    panel.grid.minor = element_blank()
  )

That’s a publication-quality plot in about 20 lines.

✏️ Exercise 4.1 — Your first ggplot

Make a scatterplot of hp (x) versus mpg (y) from cars, with points colored by cyl. Add a title.

Show solution

ggplot(cars, aes(x = hp, y = mpg, color = cyl)) +
  geom_point(size = 3) +
  labs(title = "Horsepower vs. MPG", x = "Horsepower", y = "MPG", color = "Cylinders")

✏️ Exercise 4.2 — Bars from a summary

For each transmission type (automatic / manual), compute the mean MPG. Plot it as a bar chart with the bar height equal to the mean.

Show solution

cars |>
  group_by(transmission) |>
  summarise(mean_mpg = mean(mpg), .groups = "drop") |>
  ggplot(aes(x = transmission, y = mean_mpg, fill = transmission)) +
  geom_col(width = 0.6) +
  geom_text(aes(label = round(mean_mpg, 1)), vjust = -0.4) +
  labs(title = "Mean MPG by transmission type", x = NULL, y = "Mean MPG") +
  theme_minimal() +
  theme(legend.position = "none")

✏️ Exercise 4.3 — Faceting

Make a scatterplot of wt vs. mpg, with one panel per cylinder count, and a linear fit line in each panel.

Show solution

ggplot(cars, aes(x = wt, y = mpg)) +
  geom_point() +
  geom_smooth(method = "lm", se = FALSE) +
  facet_wrap(~ cyl, labeller = label_both) +
  theme_minimal()

What’s next

You can now turn a data frame into a clear, attractive plot. Combined with Lesson 3, you’ve got a real “explore the data” workflow. Lesson 5 brings in actual statistics — t-tests, correlation, and regression — so you can quantify what your plots are showing you.

← Previous Lesson 3: Data Manipulation Next → Lesson 5: Statistics with R

Lesson 4: Data Visualization in R

Cavan Donohoe

Data Visualization in R

Setup

The ggplot2 grammar

A first plot

The most useful geoms

`geom_point()` — scatterplots

`geom_line()` — connect points by some ordering

`geom_bar()` and `geom_col()` — bar charts

`geom_histogram()` and `geom_density()` — distributions

`geom_boxplot()` — distributions by group

Faceting — small multiples

Scales — controlling axes and color

Themes — controlling the non-data ink

Saving plots

Putting it together

What’s next

Lesson 4: Data Visualization in R

Cavan Donohoe

Data Visualization in R

Setup

The ggplot2 grammar

A first plot

The most useful geoms

geom_point() — scatterplots

geom_line() — connect points by some ordering

geom_bar() and geom_col() — bar charts

geom_histogram() and geom_density() — distributions

geom_boxplot() — distributions by group

Faceting — small multiples

Scales — controlling axes and color

Themes — controlling the non-data ink

Saving plots

Putting it together

What’s next

`geom_point()` — scatterplots

`geom_line()` — connect points by some ordering

`geom_bar()` and `geom_col()` — bar charts

`geom_histogram()` and `geom_density()` — distributions

`geom_boxplot()` — distributions by group