README.md

In danielcfurr/edstan: Stan Models for Item Response Theory

edstan

The edstan package for R streamlines Bayesian estimation of item response models. It is designed around several predefined Stan models and functions that support a quick start to a Bayesian workflow.

Installation

edstan relies on the rstan package, which should be installed first. See here for instructions on installing rstan.

Once rstan is working on your machine, edstan may then be installed from CRAN:

install.packages("edstan")

Alternatively, edstan may be installed from Github:

devtools::install_github("danielcfurr/edstan")

Brief manual

A brief manual provides details on the functions and models. It also provides links to tutorials providing in depth information on each of the available models.

Quick start guide

Wide-format data

The spelling dataset has one row per person. The first column male is a dummy variable for whether the respondent is male. The remaining four columns are dummy variables for whether the respondent spelled a word correctly.

library(edstan)

# Use parallel processing
options(mc.cores = parallel::detectCores())

# Prepare data
spelling_dat <- irt_data(response_matrix = spelling[, -1])

# Fit a model to the prepared data
spelling_fit <- irt_stan(spelling_dat)

# View sampling diagnostics
stan_columns_plot(spelling_fit)

# Display a summary of parameter posteriors
print_irt_stan(spelling_fit, spelling_data)

Long-format data

The aggression dataset has one row per item response. The column person has ID values indicating the person providing the response, item has ID values for the item responded to, and poly has values 0, 1, or 2 for the item response.

library(edstan)

# Use parallel processing
options(mc.cores = parallel::detectCores())

# Prepare data
aggression_dat <- irt_data(ii = aggression$person,
                           jj = aggression$item,
                           y = aggression$poly)

# Fit a model to the prepared data
aggression_fit <- irt_stan(aggression_dat)

# View sampling diagnostics
stan_columns_plot(aggression_fit)

# Display a summary of parameter posteriors
print_irt_stan(aggression_fit, aggression_data)

danielcfurr/edstan documentation built on June 11, 2025, 6:10 p.m.