Lecture 04 — GGPlot II

New data, safe zooming, distributions, themes, and patchwork

ggplot

tidyverse

Practicing the plots you already know on a brand-new dataset, zooming safely with coord_cartesian() vs. the ylim()/scale_y_continuous(limits=) trap, histograms and density plots, a reusable custom theme file, and combining panels with patchwork.

Author

Bill Perry

Published

October 1, 2026

Where we left off (Lecture 03 — GGPlot I)

Piping straight into a plot: leaf_df %>% ggplot(...)
scale_color_manual() / scale_fill_manual() — colors you choose
facet_wrap() and facet_grid() — one panel per group
stat_summary() — mean and mean ± SE, drawn from raw data

Note

✅ Key idea from Lecture 03

You already know how to build a grouped, colored, faceted figure — but only on leaf_df. Today’s real test: can you do it on data you’ve never seen before?

Goals for Today

Practice your plotting skills on a brand-new dataset, and see where you’d find one yourself
Zoom into a plot without silently dropping data — coord_cartesian()
Build a histogram and a density plot — two views of one distribution
Apply a reusable theme file, and combine panels with patchwork

Tip

🖐 Try it yourself

By the end you will recognize the single most common ggplot mistake — zooming with ylim() / scale_y_continuous(limits = ) — before you ever make it yourself.

References:

📖 Whitlock & Schluter, Ch. 2 — Displaying Data
📖 R4DS Ch 1 — Data Visualization

Both PDFs are in the course readings/ folder.

How to Use These Slides — Predict · Type · Run

This lecture runs in five short chunks. After each chunk you switch to the activity and type the code yourself into your R script.

Note

✅ Why bother?

ylim() and coord_cartesian() produce plots that look identical for a simple scatter — the difference only shows up once you compute something (a boxplot, a smooth). Typing both yourself and comparing outputs is the only way this sticks.

🧩 Chunk 1 of 5 · New Data, Same Skills

We will cover: loading palmerpenguins, and rebuilding the two plot types you already know on data you’ve never seen.

🛑 After this chunk you will do Activity Parts 1–2.

A Dataset You Didn’t Collect

# Load all packages at the top of every script ----------
library(tidyverse)      # data wrangling + ggplot2
library(palmerpenguins) # penguins data, Palmer Station LTER

# palmerpenguins ships the data already clean -----------
penguins_df <- penguins %>%
  drop_na(species, bill_length_mm, bill_depth_mm,
          flipper_length_mm, body_mass_g)

glimpse(penguins_df)

Rows: 342
Columns: 8
$ species           <fct> Adelie, Adelie, Adelie, Adelie, Adeli…
$ island            <fct> Torgersen, Torgersen, Torgersen, Torg…
$ bill_length_mm    <dbl> 39.1, 39.5, 40.3, 36.7, 39.3, 38.9, 3…
$ bill_depth_mm     <dbl> 18.7, 17.4, 18.0, 19.3, 20.6, 17.8, 1…
$ flipper_length_mm <int> 181, 186, 195, 193, 190, 181, 195, 19…
$ body_mass_g       <int> 3750, 3800, 3250, 3450, 3650, 3625, 4…
$ sex               <fct> male, female, female, female, male, f…
$ year              <int> 2007, 2007, 2007, 2007, 2007, 2007, 2…

penguins comes inside the package — no read_excel() needed, just library(palmerpenguins)
344 penguins, 3 species, Palmer Station, Antarctica
drop_na() removes the handful of rows missing a measurement — same idea as the NAs in leaf_df

The Same Simple XY Plot — On New Data

Note

🔮 Predict first: flipper length and body mass should move together — longer flippers, heavier birds. Will three species separate into three clouds, or overlap into one?

# The exact same recipe you used on leaf_df ------------
penguins_df %>%
  ggplot(aes(x = flipper_length_mm, y = body_mass_g,
             color = species)) +
  geom_point(alpha = 0.7, size = 2) +
  labs(
    title = "Body Mass vs. Flipper Length",
    x = "Flipper Length (mm)",
    y = "Body Mass (g)",
    color = "Species"
  ) +
  theme_minimal()

Scatter plot of penguin body mass in grams against flipper length in millimeters, colored by species, showing three roughly separated clouds of points.

Same aes() + geom_point() + labs() + theme_minimal() you already typed for leaves
The dataset changed, the recipe didn’t — that’s the point
Nothing here is new syntax; it’s proof you can already do this on any data

The Same Simple Categorical Plot — On New Data

# boxplot + jitter, exactly like the leaf_df pattern ---
penguins_df %>%
  ggplot(aes(x = species, y = body_mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6, outlier.shape = NA) +
  geom_jitter(width = 0.15, alpha = 0.4) +
  labs(x = "Species", y = "Body Mass (g)",
       title = "Body Mass by Species") +
  theme_minimal() +
  theme(legend.position = "none")

Boxplot of penguin body mass in grams by species, with jittered raw points overlaid.

Tip

💡 Where would you find a dataset like this on your own?

library(palmerpenguins) or library(FSA) (fish data, Derek Ogle) — real data bundled inside an R package, zero downloading
Kaggle — search “UFO sightings” or Kaggle — search “Bigfoot sightings” — download a .csv, load it with read_csv() like any other file

Both a simple XY numeric plot and a simple categorical plot (like today’s two) work on almost any tidy dataset you find.

→ ACTIVITY 4 Parts 1–2 now

🛑 Do Activity Parts 1–2 now

Load palmerpenguins, then rebuild the XY scatter and the boxplot+jitter — same recipe, new data. Predict, type, run.

🧩 Chunk 2 of 5 · Zooming In Safely

We will cover: coord_cartesian() to zoom — and the ylim() / scale_y_continuous(limits = ) trap to avoid.

🛑 After this chunk you will do Activity Part 3.

`coord_cartesian()` — Zoom the View, Keep the Data

# coord_cartesian() zooms the VIEW, keeps all the data -
penguins_df %>%
  ggplot(aes(x = species, y = body_mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6, outlier.shape = NA) +
  coord_cartesian(ylim = c(3000, 5000)) +
  labs(x = "Species", y = "Body Mass (g)",
       title = "Zoomed with coord_cartesian()") +
  theme_minimal() +
  theme(legend.position = "none")

Boxplot of penguin body mass by species zoomed to a narrower y-axis window using coord_cartesian, with all 342 penguins still contributing to each box.

Zooms the view — every penguin still contributes to each box’s median and quartiles
The window just crops what you can see afterward
This is the one to reach for whenever you just want a closer look

The Trap — `ylim()` / `scale_y_continuous(limits = )` Delete Rows First

Note

🔮 Predict first: Gentoo penguins run heavier than the other two species — many sit above 5000 g. If we drop every penguin outside 3000–5000 g before drawing the boxes, what happens to the Gentoo box specifically?

# ylim() DELETES rows outside the range first ----------
penguins_df %>%
  ggplot(aes(x = species, y = body_mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6, outlier.shape = NA) +
  ylim(3000, 5000) +
  labs(x = "Species", y = "Body Mass (g)",
       title = "ylim() — boxes recomputed on fewer penguins") +
  theme_minimal() +
  theme(legend.position = "none")

Boxplot of penguin body mass by species using ylim, with the Gentoo box recomputed from only the penguins remaining after out-of-range rows were silently dropped.

Warning

⚠️ Watch out!

ylim() — and its long form scale_y_continuous(limits = c(3000, 5000)) — silently delete every row outside the range before anything is computed. Compare the Gentoo box above to the coord_cartesian() version: same window, different box.

Rule: zoom with coord_cartesian(). Only use ylim() / scale_y_continuous(limits = ) when you actually want those rows gone from the analysis.

→ ACTIVITY 4 Part 3 now

🛑 Do Activity Part 3 now

Zoom with coord_cartesian(), then run the ylim() version and count exactly how many penguins it drops. Predict, type, run.

🧩 Chunk 3 of 5 · Two Views of One Distribution

We will cover: geom_histogram() and geom_density() — the two standard ways to show how one numeric variable is distributed.

🛑 After this chunk you will do Activity Part 4.

`geom_histogram()` — Bars of Counts

# geom_histogram() bins the data and counts rows per bin
penguins_df %>%
  ggplot(aes(x = body_mass_g)) +
  geom_histogram(binwidth = 200, fill = "steelblue",
                 color = "white") +
  labs(x = "Body Mass (g)", y = "Count",
       title = "Distribution of Body Mass") +
  theme_minimal()

Histogram of penguin body mass in grams, showing a somewhat bimodal distribution.

binwidth sets the width of each bar in the units of x — here, 200 g per bin
Too narrow → noisy, spiky bars. Too wide → real structure gets smoothed away
Try a few binwidth values before settling on one

`geom_density()` — a Smooth Curve Instead of Bars

# geom_density() smooths the same shape into a curve ---
penguins_df %>%
  ggplot(aes(x = body_mass_g)) +
  geom_density(fill = "steelblue", alpha = 0.5) +
  labs(x = "Body Mass (g)", y = "Density",
       title = "Density of Body Mass") +
  theme_minimal()

Smoothed density curve of penguin body mass in grams, covering the same range as the histogram above.

Same variable, same shape — a density curve is a smoothed histogram, area under the curve sums to 1
The y-axis is no longer a count — it’s a density, which is why the numbers look small
📖 Whitlock & Schluter, Ch. 2 — Displaying Data

Tip

Coming soon: in the Summary Statistics lecture we’ll come back to this exact curve and shade in the 95% range underneath it. Today, just get comfortable building both plot types.

→ ACTIVITY 4 Part 4 now

🛑 Do Activity Part 4 now

Build a histogram and a density plot of the same variable. Predict, type, run.

🧩 Chunk 4 of 5 · A Reusable Theme File

We will cover: sourcing a custom theme file and applying it — no need to build one from scratch today.

🛑 After this chunk you will do Activity Part 5.

`source()` a Theme File Instead of Repeating `theme()`

# source() loads functions from another .R file -------
source("themes/r_themes_for_3_sizes.R")

penguins_df %>%
  ggplot(aes(x = species, y = body_mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6, outlier.shape = NA) +
  labs(x = "Species", y = "Body Mass (g)",
       title = "Body Mass by Species") +
  theme_regular() +
  theme(legend.position = "none")

Boxplot of penguin body mass by species rendered with the custom theme_regular function sourced from an external file, with a clean white background and bold axis titles.

source("themes/r_themes_for_3_sizes.R") runs that file once and loads theme_small(), theme_regular(), theme_large() into your session
You don’t need to understand every line inside the file today — just how to call it
Pick the size that matches where the figure is going: slides/report → theme_regular()

Tweaking Fonts and Colors on Top

# theme() and scale_fill_manual() still work as usual --
penguins_df %>%
  ggplot(aes(x = species, y = body_mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6, outlier.shape = NA) +
  scale_fill_manual(values = c(
    "Adelie" = "darkorange", "Chinstrap" = "purple",
    "Gentoo" = "cyan4"
  )) +
  labs(x = "Species", y = "Body Mass (g)",
       title = "Body Mass by Species") +
  theme_regular() +
  theme(
    legend.position = "none",
    plot.title = element_text(size = rel(1.4), face = "bold")
  )

The same boxplot with a larger bold title and custom fill colors layered on top of the sourced theme.

A sourced theme is still just a theme() call — + theme(...) after it overrides individual pieces, same as any built-in theme
scale_fill_manual() still controls colors; the theme controls everything else (fonts, lines, margins)

Note

✅ Key idea

One theme file, sourced at the top of every script, keeps every figure in a project looking consistent — change the file once, every plot updates.

→ ACTIVITY 4 Part 5 now

🛑 Do Activity Part 5 now

Download the theme file, source it, and apply theme_regular() to one of your plots. Then tweak a font size and a color. Predict, type, run.

🧩 Chunk 5 of 5 · Combining Panels with `patchwork`

We will cover: patchwork — putting two ggplot objects side by side as one figure.

🛑 After this chunk you will do Activity Part 6.

`patchwork` — Two Plots, One Figure

library(patchwork) # combine ggplot objects with + and /

xy_plot <- penguins_df %>%
  ggplot(aes(x = flipper_length_mm, y = body_mass_g,
             color = species)) +
  geom_point(alpha = 0.7) +
  labs(x = "Flipper Length (mm)", y = "Body Mass (g)",
       color = "Species") +
  theme_regular()

box_plot <- penguins_df %>%
  ggplot(aes(x = species, y = body_mass_g, fill = species)) +
  geom_boxplot(alpha = 0.6, outlier.shape = NA) +
  labs(x = "Species", y = "Body Mass (g)") +
  theme_regular() +
  theme(legend.position = "none")

xy_plot + box_plot +
  plot_annotation(
    title = "Penguin Body Mass, Two Ways",
    tag_levels = "A"
  )

Two panels combined side by side with patchwork: a scatter plot of body mass against flipper length on the left, and a boxplot of body mass by species on the right, with the overall figure labeled panel A and B.

Each panel is an ordinary saved ggplot object — build them separately first, exactly like you already do
+ places two patchwork-ready plots side by side; / stacks them
plot_annotation() adds one title and A/B/C tags across the whole combined figure, not any single panel

→ ACTIVITY 4 Part 6 now

🛑 Go to Activity 4 — Part 6

Close the slides. Combine two of your plots with patchwork and plot_annotation(), then save the combined figure.

What We Learned Today

The XY-scatter / boxplot+jitter recipe you already know works on any tidy dataset — palmerpenguins, FSA, or a .csv from Kaggle
coord_cartesian() zooms safely; ylim() / scale_y_continuous(limits = ) silently deletes rows first
geom_histogram() and geom_density() — two views of one distribution
source() a theme file once, then call theme_regular() (or _small() / _large()) everywhere
patchwork (+ / /) combines saved plots; plot_annotation() titles and tags the whole figure

References:

📖 Whitlock & Schluter, Ch. 2 — Displaying Data
📖 R4DS Ch 1 — Data Visualization

Up next — Lecture 05, Wrangling:

filter(), select(), mutate(), arrange()
Chaining verbs into one pipeline

Other Formats

Where we left off (Lecture 03 — GGPlot I)

Goals for Today

How to Use These Slides — Predict · Type · Run

🧩 Chunk 1 of 5 · New Data, Same Skills

A Dataset You Didn’t Collect

The Same Simple XY Plot — On New Data

The Same Simple Categorical Plot — On New Data

→ ACTIVITY 4 Parts 1–2 now

🧩 Chunk 2 of 5 · Zooming In Safely

coord_cartesian() — Zoom the View, Keep the Data

The Trap — ylim() / scale_y_continuous(limits = ) Delete Rows First

→ ACTIVITY 4 Part 3 now

🧩 Chunk 3 of 5 · Two Views of One Distribution

geom_histogram() — Bars of Counts

geom_density() — a Smooth Curve Instead of Bars

→ ACTIVITY 4 Part 4 now

🧩 Chunk 4 of 5 · A Reusable Theme File

source() a Theme File Instead of Repeating theme()

Tweaking Fonts and Colors on Top

→ ACTIVITY 4 Part 5 now

🧩 Chunk 5 of 5 · Combining Panels with patchwork

patchwork — Two Plots, One Figure

→ ACTIVITY 4 Part 6 now

What We Learned Today

`coord_cartesian()` — Zoom the View, Keep the Data

The Trap — `ylim()` / `scale_y_continuous(limits = )` Delete Rows First

`geom_histogram()` — Bars of Counts

`geom_density()` — a Smooth Curve Instead of Bars

`source()` a Theme File Instead of Repeating `theme()`

🧩 Chunk 5 of 5 · Combining Panels with `patchwork`

`patchwork` — Two Plots, One Figure