https://github.com/wardbrian/scikit-stan

A sklearn style interface to Stan regression models
https://github.com/wardbrian/scikit-stan

python scikit-learn sklearn-compatible stan

Last synced: 3 months ago
JSON representation

A sklearn style interface to Stan regression models

• Host: GitHub
• URL: https://github.com/wardbrian/scikit-stan
• Owner: WardBrian
• License: bsd-3-clause
• Created: 2022-06-06T15:47:00.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-10-14T01:26:51.000Z (3 months ago)
• Last Synced: 2024-10-14T19:21:02.032Z (3 months ago)
• Topics: python, scikit-learn, sklearn-compatible, stan
• Language: Python
• Homepage: https://scikit-stan.readthedocs.io/
• Size: 9.48 MB
• Stars: 25
• Watchers: 3
• Forks: 0
• Open Issues: 13
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# scikit-stan

Scikit-Stan is a package of [Stan](https://mc-stan.org/) models wrapped in a

[Scikit-Learn](https://scikit-learn.org/stable/) style interface.

This package is currently under active development and should be treated as beta software.

Documentation is available at https://brianward.dev/scikit-stan/ or on [ReadTheDocs](https://scikit-stan.readthedocs.io/)

(older versions and PDFs available).

## Installation

Pre-compiled wheels for the package are available for MacOS, Windows, and Linux systems via `pip install scikit_stan`.

Source installation requires a working installation of [CmdStan](https://mc-stan.org/docs/cmdstan-guide/index.html).

## Basic usage

```python

from scikit_stan import GLM

m = GLM(family='gamma') # Gamma family distribution with canonical inverse link

m.fit(X, y) # runs HMC-NUTS

predictions = m.predict(X) # generates new predictions from fitted model

score = m.score(X, y) # computes the R2 score of the fitted model on the data X and observations y

```

## Attribution

This package is licensed under the BSD 3-clause license.

It is inspired by existing packages in the Stan ecosystem like

[rstanarm](https://github.com/stan-dev/rstanarm).

This package was initially developed at the Simons Foundation by Alexey Izmailov during

a summer 2022 internship under the mentorship of Brian Ward.