Categorical Data

Descriptive summaries & visualizations (freq, proportion, crosstabs, bar/pie/mosaic/waffle)

1 Introduction

  • Recognize and work with categorical variables in R.
  • Summarize categories using frequencies and proportions (a.k.a. relative frequency).
  • Create standard plots for categorical data: bar, stacked bar, pie, mosaic, and waffle.K
  • Read and interpret cross‑tabulations (two‑way tables) with row/column/table proportions.
Tip

Use the Copy button on each code chunk. Many chunks include a template version followed by a worked example.

2 Google Colab Setup

Copy the following code and put it in a code cell in Google Colab. Only do this if you are using a completely new notebook.

# This code will load the R packages we will use
install.packages(c("rcistats"),
                 repos = c("https://inqs909.r-universe.dev", "https://cloud.r-project.org"))
library(rcistats)
library(tidyverse)


# Uncomment and run for themes
# rcistats::install_themes()
# library(ThemePark)
# library(ggthemes)

2.1 Data for this handout

We will use the Great American Coffee Taste Test survey data from TidyTuesday. Below is a subset of the data.

Code 
coffee <- read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2024/2024-05-14/coffee_survey.csv")

2.2 Using the templates: what to change

Use this legend whenever you see a Template code block.

  • DATA → replace with your data frame/tibble name (e.g., coffee).
  • VAR → replace with the single categorical variable you want (e.g., caffeine).
    • In ggplot(DATA, aes(x = VAR)), write ggplot(coffee, aes(x = caffeine)).
    • In functions that take a vector (e.g., cat_stats(DATA$VAR)), write cat_stats(coffee$caffeine).
  • VAR1 and VAR2 → replace with the first and second categorical variables for two‑way tables/stacked bars (e.g., caffeine, taste).
  • DF / wdf / df_pie / waffle_df → these are intermediate objects created in the chunk. You can keep the same names or rename them; if you rename, update the subsequent line that uses them.
  • group = 1 → keep this as‑is for one‑variable proportion bar charts; it ensures correct normalization.

2.2.1 Quick replace checklist

  1. Swap DATA for your data frame (usually coffee).
  2. Swap VAR for your categorical column (e.g., caffeine).
  3. For two‑variable templates, set VAR1 and VAR2 (e.g., caffeine and taste).
  4. If you change any intermediate object name (like df_pie), update it on the next line as well.

2.2.2 Tiny example

Template

# Frequency bar (template)
 ggplot(DATA, aes(x = VAR)) +
   geom_bar()

Filled‑in

# Frequency bar (coffee example)
 ggplot(coffee, aes(x = caffeine)) +
   geom_bar()

Template

# Crosstab row proportions (template)
 cat_stats(VAR1, VAR2, prop = "row")

Filled‑in

# Crosstab row proportions (coffee example)
 cat_stats(coffee$caffeine, coffee$taste, prop = "row")

3 Categorical Data

Categorical data record membership in a set of categories (levels), e.g., “Yes/No”, “Major”, or “City”.

  • Stored as text (character/factor) or as codes like 1, 2, 3 with a codebook describing the labels.

4 One‑variable summaries

4.1 Frequency (counts)

Definition: number of observations in each category.

Template:

# Replace DATA$VAR with your variable
# freq table (counts)
cat_stats(DATA$VAR)

Example: caffeine preference (coffee$caffeine)

cat_stats(coffee$caffeine)
#> Continguency Table 
#>  
#>                  n   prop
#> Decaf          136 0.0347
#> Full caffeine 3576 0.9129
#> Half caff      205 0.0523
#> 
#> Number of Missing: 125
#> Proportion of Missing: 0.0309252845126175
#> Row Variable: coffee$caffeine

4.2 Proportion (relative frequency)

Definition: share of the sample in each category; comparable across sample sizes.

Template:

# proportions only
cat_stats(DATA$VAR)

Example:

cat_stats(coffee$caffeine)
#> Continguency Table 
#>  
#>                  n   prop
#> Decaf          136 0.0347
#> Full caffeine 3576 0.9129
#> Half caff      205 0.0523
#> 
#> Number of Missing: 125
#> Proportion of Missing: 0.0309252845126175
#> Row Variable: coffee$caffeine

5 Two‑variable summaries (cross‑tabulation)

5.1 Cross‑tabulation (two‑way table)

  • Rows: categories of one variable
  • Columns: categories of the second variable
  • Report counts or proportions by table, row, or column

ALL:

cat_stats(coffee$caffeine, coffee$taste)
#> Continguency Table 
#>  
#>               No                              Yes                              
#> Decaf         "26 / 0.0073 / 0.208 / 0.2653"  "99 / 0.0279 / 0.792 / 0.0287"   
#> Full caffeine "66 / 0.0186 / 0.0203 / 0.6735" "3178 / 0.8957 / 0.9797 / 0.9212"
#> Half caff     "6 / 0.0017 / 0.0335 / 0.0612"  "173 / 0.0488 / 0.9665 / 0.0501" 
#> Col Totals    "98 / 0.0276"                   "3450 / 0.9724"                  
#>               Row Totals     
#> Decaf         "125 / 0.0352" 
#> Full caffeine "3244 / 0.9143"
#> Half caff     "179 / 0.0505" 
#> Col Totals    "Total: 3548"  
#> 
#> Cell Contents: n / tbl % / row % / col % 
#> Col Totals Contents: n / row % 
#> Row Totals Contents: n / col % 
#> Column Variable: coffee$taste
#> Row Variable: coffee$caffeine

Table proportions (each cell ÷ grand total):

cat_stats(coffee$caffeine, coffee$taste, prop = "table")
#> Continguency Table 
#>  
#>               No            Yes             Row Totals     
#> Decaf         "26 / 0.0073" "99 / 0.0279"   "125 / 0.0352" 
#> Full caffeine "66 / 0.0186" "3178 / 0.8957" "3244 / 0.9143"
#> Half caff     "6 / 0.0017"  "173 / 0.0488"  "179 / 0.0505" 
#> Col Totals    "98 / 0.0276" "3450 / 0.9724" "Total: 3548"  
#> 
#> Cell Contents: n / tbl % 
#> Col Totals Contents: n / row % 
#> Row Totals Contents: n / col % 
#> Column Variable: coffee$taste
#> Row Variable: coffee$caffeine

Row proportions (each cell ÷ its row total):

cat_stats(coffee$caffeine, coffee$taste, prop = "row")
#> Continguency Table 
#>  
#>               No            Yes             Row Totals     
#> Decaf         "26 / 0.208"  "99 / 0.792"    "125 / 0.0352" 
#> Full caffeine "66 / 0.0203" "3178 / 0.9797" "3244 / 0.9143"
#> Half caff     "6 / 0.0335"  "173 / 0.9665"  "179 / 0.0505" 
#> Col Totals    "98 / 0.0276" "3450 / 0.9724" "Total: 3548"  
#> 
#> Cell Contents: n / row % 
#> Col Totals Contents: n / row % 
#> Row Totals Contents: n / col % 
#> Column Variable: coffee$taste
#> Row Variable: coffee$caffeine

Column proportions (each cell ÷ its column total):

cat_stats(coffee$caffeine, coffee$taste, prop = "col")
#> Continguency Table 
#>  
#>               No            Yes             Row Totals     
#> Decaf         "26 / 0.2653" "99 / 0.0287"   "125 / 0.0352" 
#> Full caffeine "66 / 0.6735" "3178 / 0.9212" "3244 / 0.9143"
#> Half caff     "6 / 0.0612"  "173 / 0.0501"  "179 / 0.0505" 
#> Col Totals    "98 / 0.0276" "3450 / 0.9724" "Total: 3548"  
#> 
#> Cell Contents: n / col % 
#> Col Totals Contents: n / row % 
#> Row Totals Contents: n / col % 
#> Column Variable: coffee$taste
#> Row Variable: coffee$caffeine

6 Visualizing Categorical Data

6.1 Bar plots

6.1.1 Frequency bar plot

Template (frequency):

# ggplot() + geom_bar() counts rows per category by default
ggplot(DATA, aes(x = VAR)) +
  geom_bar()

Example:

ggplot(coffee, aes(x = caffeine)) +
  geom_bar()

6.1.2 Relative frequency bar plot

Template (proportion):

# after_stat(prop) computes proportions within the layer
ggplot(DATA, aes(x = VAR, y = after_stat(prop), group = 1)) +
  geom_bar()

Example:

ggplot(coffee, aes(x = caffeine, y = after_stat(prop), group = 1)) +
  geom_bar()

Note

Tip: Add labels/theme as needed: labs(x = "", y = "Proportion") + theme_minimal()

6.2 Stacked bar plots

Template:

ggplot(DATA, aes(x = VAR1, fill = VAR2)) +
  geom_bar()

Example (stacked counts):

ggplot(coffee, aes(x = caffeine, fill = taste)) +
  geom_bar()

A bar plot showing the caffeine levels by different tastes.

Example (horizontal):

ggplot(coffee, aes(y = caffeine, fill = taste)) +
  geom_bar()

Template (stacked proportions):

ggplot(DATA, aes(x = VAR1, fill = VAR2)) +
  geom_bar(position = "fill") +
  labs(y = "Proportion")

6.3 Pie charts (use sparingly)

Note: Pie charts can be harder to compare precisely than bars. If you use them, label clearly.

Template:

df_pie <- cat_stats(DATA$VAR, tbl_df = TRUE)$table
ggplot(df_pie, aes(cat = Category, val = n, fill = Category)) +
  geom_pie()

Example:

coffee_pie <- cat_stats(coffee$caffeine, tbl_df = TRUE)$table
ggplot(coffee_pie, aes(cat = Category, val = n, fill = Category)) +
  geom_pie()

7 Appendix: minimal templates (copy‑paste)

Each template below has placeholders in ALL CAPS (e.g., DATA, VAR, VAR1, VAR2). Replace them with your own dataset name and variable names.

7.1 Frequency table

cat_stats(DATA$VAR)
  • DATA → the name of your dataset (e.g., coffee).
  • VAR → a single categorical variable (e.g., caffeine).

7.2 Proportions only

cat_stats(DATA$VAR, prop_only = TRUE)
  • DATA → the name of your dataset (e.g., coffee).
  • VAR → a single categorical variable (e.g., caffeine).

7.3 Bar Plot: Frequency

ggplot(DATA, aes(x = VAR)) + 
    geom_bar()
  • DATA → the name of your dataset (e.g., coffee).
  • VAR → a single categorical variable (e.g., caffeine).

7.4 Bar Plot: Proportion

ggplot(DATA, aes(x = VAR, y = after_stat(prop), group = 1)) + 
    geom_bar()
  • DATA → the name of your dataset (e.g., coffee).
  • VAR → a single categorical variable (e.g., caffeine).

7.5 Cross-tabulations (all)

cat_stats(DATA$VAR1, DATA$VAR2)
  • DATA → the name of your dataset (e.g., coffee).
  • VAR1 → a single categorical variable (e.g., caffeine).
  • VAR2 → a single categorical variable (e.g., taste).

7.6 Cross-tabulations (proportions)

cat_stats(DATA$VAR1, DATA$VAR2, prop = "table")
cat_stats(DATA$VAR1, DATA$VAR2, prop = "row")
cat_stats(DATA$VAR1, DATA$VAR2, prop = "col")
  • DATA → the name of your dataset (e.g., coffee).
  • VAR1 → a single categorical variable (e.g., caffeine).
  • VAR2 → a single categorical variable (e.g., taste).

7.7 Stacked bar (counts)

ggplot(DATA, aes(x = VAR1, fill = VAR2)) + geom_bar()
  • DATA → the name of your dataset (e.g., coffee).
  • VAR1 → a single categorical variable (e.g., caffeine).
  • VAR2 → a single categorical variable (e.g., taste).

7.8 Stacked bar (proportions)

ggplot(DATA, aes(x = VAR1, fill = VAR2)) +
  geom_bar(position = "fill") + labs(y = "Proportion")
  • DATA → the name of your dataset (e.g., coffee).
  • VAR1 → a single categorical variable (e.g., caffeine).
  • VAR2 → a single categorical variable (e.g., taste).

7.9 Pie Chart

df_pie <- cat_stats(DATA$VAR, tbl_df = TRUE)$table
ggplot(df_pie, aes(cat = Category, val = n, fill = Category)) + 
    geom_pie()
  • DATA → the name of your dataset (e.g., coffee).
  • VAR → a single categorical variable (e.g., caffeine).