Grammar of data visualization

Lecture 3

Dr. Mine Çetinkaya-Rundel

Duke University
STA 199 - Fall 2026

August 31, 2026

Warm up

While you wait…

1. Clone your ae repository

Go to the course GitHub organization and clone ae-YOUR-GITHUB-NAME repo to your Positron session in your container.

2. Participate 📱💻

Remember this visualization from the code along video – what was it about?

QR code for Wooclap

Go to wooclap.com and use the code IOAADOR.

Outline

  • Last week:

    • We introduced you to the course toolkit.

    • You cloned your lab repository and started making some updates in your Quarto documents.

    • You committed and pushed your changes back – at least most of you did!

  • Today:

    • You will clone your ae (application exercise) repository.

    • We will introduce data visualization.

    • You will work on the application exercise on data visualization, commit your changes, and push them.

Data visualization

Load packages

library(tidyverse)


Reminder

tidyverse is the collection of R packages designed for data science.

Load the data

gss <- read_csv("data/gss-2024.csv")


Reminder

We can read this as “read the CSV file called gss-2024.csv in the data folder” and save the result as gss.

2024 General Social Survey (GSS)

GSS is a widely respected, nationally representative sociological survey that tracks how American opinions, attitudes, and behaviors change over time.

We’re working with GSS data from the most recent wave:

  • Each row represents one survey respondent.
  • The variables describe demographics, education, family, happiness, employment, and political identification.

Note

The GSS is a cross-sectional survey. We can describe associations, but these data do not establish causation.

View the data

gss
# A tibble: 3,294 × 9
    year marital_status   age education_years children happiness     health
   <dbl> <chr>          <dbl>           <dbl>    <dbl> <chr>         <chr> 
 1  2024 Never married     33              16        2 Not too happy Good  
 2  2024 Never married     64              16        0 Pretty happy  Good  
 3  2024 Married           69              14        0 Pretty happy  Good  
 4  2024 Never married     19              12        0 Pretty happy  Good  
 5  2024 Divorced          70              13        3 Not too happy Good  
 6  2024 Married           53              14        2 Pretty happy  Good  
 7  2024 Married           48              13        6 Pretty happy  Good  
 8  2024 Divorced          30              14        1 Not too happy Good  
 9  2024 Married           60              14        2 Very happy    Good  
10  2024 Never married     25              12        0 Pretty happy  Good  
# ℹ 3,284 more rows
# ℹ 2 more variables: employment_status <chr>, party_id <chr>

Variables in our extract

Variable Description Type
marital_status Current marital status Categorical
age Age in years Numerical
education_years Completed years of education Numerical
children Number of children; 8 means 8 or more Numerical
happiness Self-reported happiness Categorical
health Self-reported general health Categorical
employment_status Current employment status Categorical
party_id Party identification Categorical

Start with a question

How does age vary across marital-status groups?

Before writing code:

  • Which variables do we need?
  • What type is each variable?
  • What kind of plot might answer the question?

Age by marital status

ggplot(gss, aes(x = marital_status, y = age, fill = marital_status)) +
  geom_boxplot(show.legend = FALSE) +
  labs(
    title = "Age varies substantially across marital-status groups",
    x = "Marital status",
    y = "Age (years)",
    caption = "Source: 2024 General Social Survey (GSS), NORC"
  ) +
  scale_fill_viridis_d()
Warning: Removed 94 rows containing non-finite outside the scale range
(`stat_boxplot()`).

Participate 📱💻

The plot was created, but R displayed the warning below. What happened to the 94 observations?

Warning: Removed 94 rows containing non-finite values
outside the scale range (`stat_boxplot()`).

QR code for Wooclap

Go to wooclap.com and use the code IOAADOR.

Interpret the plot

With a partner:

  • Which group has the highest median age? The lowest?
  • Which groups have the greatest variability?
  • What does one box summarize?
  • Can this plot tell us whether marital status causes differences in age?

Plot autopsy

ggplot(
1  data = gss,
2  mapping = aes(x = marital_status, y = age, fill = marital_status)
) +
3  geom_boxplot() +
4  labs(...) +
5  scale_fill_viridis_d()
1
The data to visualize
2
The variables mapped to visual properties
3
The geometric object used to represent the observations
4
The plot labels
5
The scale used to map values to colors

Which component would you change to plot different variables? Which component controls the variable represented by the fill color?

Mapping and aesthetics

Aesthetics are visual properties such as position, color, fill, shape, size, and transparency.

For the GSS boxplot:

aes(
  x = marital_status,
  y = age,
  fill = marital_status
)
Variable Aesthetic
marital_status x position
age y position
marital_status fill color

Map a variable; set a value

Map: fill varies with a variable, so ggplot2 creates a legend

geom_boxplot(aes(fill = marital_status))

Set: every box has the same fill, so no legend is needed

geom_boxplot(fill = "pink")

Tip

Variables belong inside aes(); fixed visual choices belong outside it.

Grammar of graphics

  • When building or reading a plot, work through the layers as if they’re a checklist.

  • Not every plot needs every component, but each component is a separate decision.

Application exercise

Let’s get laptops out!