regdatasets

regdatasets provides a curated collection of 24 datasets for teaching introductory and intermediate regression analysis in education and social science contexts.

Installation

Install the development version from GitHub:

# install.packages("devtools")
devtools::install_github("joonho112/regdatasets")

Datasets

Dataset	N	Variables	Topic
gcse	4,059	9	GCSE exam scores and London Reading Test (Chapters 1–2, 5)
crime	67	5	Florida county crime rates (Chapters 4–5)
instruction	1,190	12	Math achievement gains, multilevel (Chapters 2–3)
reading	66	8	Three-group reading instruction experiment (Chapter 3)
pisa2000	4,528	15	OECD PISA reading scores, 3 countries (Chapter 1)
individuals	55,899	6	BLS wage data by gender and sector (Chapter 5)
faculty	514	10	Faculty salaries and experience (Chapter 5)
nels_data	2,000	299	National Education Longitudinal Study (Chapters 7–8)
hsbs1	350	16	High School and Beyond survey (Chapter 8)
gss_1	2,832	16	General Social Survey 1982/1994 (Chapter 9)
berkeley	3,593	3	UC Berkeley admissions, Simpson’s paradox (Chapter 10)
penalty	674	3	Death penalty sentencing by race (Chapter 11)
titanic	1,309	6	Titanic passenger survival (Chapters 9–11)
womenlf	263	5	Women’s labor force participation (Chapter 13)
satisfaction	110	3	Ordinal satisfaction survey (Chapter 13)
alcohol1_pp	246	9	Adolescent alcohol use, person-period (Chapter 13)
lambert	492	109	Canadian longitudinal child study (Chapter 12)
civic_ed	794	237	Civic education pre/post study (Chapters 1, 7)
disc	2,000	15	NELS:88 discipline study (Chapters 9–10)
disc2	2,000	15	NELS:88 with achievement scores (Chapters 10–11)
berk_sub	1,169	3	Berkeley admissions subset (Chapter 10)
hsb_sub	188	7	High School and Beyond subset (Chapter 7)
classdata_07	44	12	Classroom survey data (Chapter 1)
grades	198	8	Essay grades and writing features (Chapter 12)

Quick Start

library(regdatasets)

# --- Simple linear regression ---
data(gcse)
model <- lm(gcse ~ lrt, data = gcse)
summary(model)

# --- Logistic regression: Simpson's paradox ---
data(berkeley)
# Unadjusted: gender appears to matter
crude <- glm(admitted ~ female, family = binomial, data = berkeley)
# Adjusted: gender effect disappears after controlling for department
adjusted <- glm(admitted ~ female + factor(department),
                family = binomial, data = berkeley)
exp(coef(crude)["female"])    # Crude OR ~ 0.64
exp(coef(adjusted)["female"]) # Adjusted OR ~ 0.97

# --- One-way ANOVA ---
data(reading)
model_anova <- lm(post1 ~ factor(group), data = reading)
anova(model_anova)

Works Well With

regdatasets is designed to work seamlessly with modern R workflows:

library(tidyverse)   # dplyr, ggplot2, tibble, ...
library(easystats)   # parameters, performance, effectsize, see, ...
library(gtsummary)   # publication-ready tables
library(marginaleffects) # predictions, comparisons, slopes

Example using the easystats and marginaleffects ecosystem:

library(regdatasets)
library(easystats)
library(marginaleffects)

# Fit model
data(gcse)
model <- lm(gcse ~ lrt * school, data = gcse)

# easystats: model summary and diagnostics
model_parameters(model)
check_model(model)

# marginaleffects: predicted values and slopes
plot_predictions(model, condition = c("lrt", "school"))
avg_slopes(model, variables = "lrt", by = "school")

Course Coverage

Chapters	Topic	Datasets
1–2	SLR, ANCOVA	gcse, civic_ed, classdata_07, pisa2000
3	One-way ANOVA	reading, instruction
4	Continuous predictors	crime
5	Interactions	gcse, individuals, faculty, crime
6	Nonlinear relationships	nels_data, crime
7–8	Model building & diagnostics	nels_data, hsbs1, hsb_sub
9–11	Logistic regression	gss_1, berkeley, penalty, titanic, disc
12	Latent response & GLM	lambert, grades
13	Ordinal models	womenlf, satisfaction, alcohol1_pp

Companion Textbook

This package accompanies Regression Lecture Notes, a Quarto-based interactive textbook for graduate-level regression analysis in education and social science.

License

MIT