Probabilistically-autoencoded horseshoe-disentangled multidomain item-response theory models

Update as of June 2026

This repository has been archived and is no longer maintained. The code is provided for historical reference and may contain unpatched or unknown vulnerabilities. It should not be used in production systems.

Background

This code aims to factor item response data using the graded response model and sparse coding.

Check out these two manuscripts for details:

• https://arxiv.org/abs/1912.02351
• https://proceedings.mlr.press/v206/chang23a.html

Installation

Using pip:

pip install git+https://github.com/CC-RMD-EpiBio/autoencirt.git

Examples

Here is an example training notebook:

Factorization of the Right-Wing-Authoritarianism Scale