A Python package for computerized adaptive testing (CAT). It provides a reusable CAT engine, simulation tools, and multiple methods for:
- test initialization (selecting the initial ability of the examinees)
- item selection
- ability estimation
- test stopping
Installation
Install it using pip install catsim.
Basic Usage
Tutorial notebooks are available in the repository under notebooks/:
- Introduction
- Manual CAT sessions
- Simulation studies
- Item selection and stopping
- Visualization and analysis
- Have an item matrix;
- Have a sample of examinee proficiencies, or a number of examinees to be generated;
- Create an initializer,
- Create a simulation runner
- Run the simulation and inspect the returned result object;
- Plot your results.
from catsim.initialization import RandomInitializer
from catsim.selection import MaxInfoSelector
from catsim.estimation import NumericalSearchEstimator
from catsim.stopping import MinErrorStopper
from catsim.simulation import SimulationRunner
from catsim import ItemBank
initializer = RandomInitializer()
selector = MaxInfoSelector()
estimator = NumericalSearchEstimator()
stopper = MinErrorStopper(0.4, max_items=20)
runner = SimulationRunner(ItemBank.generate_item_bank(100), initializer, selector, estimator, stopper)
result = runner.run(10)
print(result.rmse)Dependencies
Runtime dependencies are declared in pyproject.toml and are installed automatically with pip install catsim or uv pip install catsim.
If you are contributing from a clone of the repository, install the full developer toolchain (tests, linting, type checking, notebooks, coverage, mutation testing) with:
uv sync --group devTo build the HTML documentation locally, install Sphinx and its extensions with:
uv sync --group docsCompatibility
_catsim_ is compatible and tested against Python 3.11, 3.12, 3.13, and 3.14.
Important links
- Official source code repo:
- HTML documentation (stable release):
- Issue tracker:
Citing catsim
You can cite the package using the following bibtex entry:
@article{catsim,
author = {Meneghetti, Douglas De Rizzo and Aquino Junior, Plinio Thomaz},
title = {Application and simulation of computerized adaptive tests through the package catsim},
year = 2018,
month = jul,
archiveprefix = {arXiv},
eprint = {1707.03012},
eprinttype = {arxiv},
journal = {arXiv:1707.03012 [stat]},
primaryclass = {stat}
}What's a CAT
Computerized adaptive tests are educational evaluations, usually taken by examinees in a computer or some other digital means, in which the examinee's ability is evaluated after the response of each item. The new ability is then used to select a new item, closer to the examinee's real ability. This method of test application has several advantages compared to the traditional paper-and-pencil method or even linear tests applied electronically, since high-ability examinees are not required to answer all the easy items in a test, answering only the items that actually give some information regarding his or hers true knowledge of the subject at matter. A similar, but inverse effect happens for those examinees of low ability level.
More information is available in the docs and over at Wikipedia.
If you are looking for IRT item parameter estimation
_catsim_ does not implement item parameter estimation. I have had great joy outsourcing that functionality to the mirt R package along the years. However, since many users request packages with item parameter estimation capabilities in the Python ecosystem, here are a few links. While I have not used them personally, specialized packages like these are hard to come by, so I hope these are helpful.