https://github.com/eigenfoo/tests-as-linear

Python port of "Common statistical tests are linear models" by Jonas Kristoffer Lindeløv.
https://github.com/eigenfoo/tests-as-linear

linear-models python scipy statistical-tests statsmodels

Last synced: 12 days ago
JSON representation

Python port of "Common statistical tests are linear models" by Jonas Kristoffer Lindeløv.

• Host: GitHub
• URL: https://github.com/eigenfoo/tests-as-linear
• Owner: eigenfoo
• License: cc-by-4.0
• Created: 2019-06-23T08:59:28.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2024-08-21T00:17:21.000Z (3 months ago)
• Last Synced: 2024-10-15T01:21:33.822Z (28 days ago)
• Topics: linear-models, python, scipy, statistical-tests, statsmodels
• Language: HTML
• Homepage: https://eigenfoo.xyz/tests-as-linear/
• Size: 4.93 MB
• Stars: 88
• Watchers: 6
• Forks: 23
• Open Issues: 7
• Metadata Files:
  • Readme: README.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE.txt
  • Code of conduct: CODE_OF_CONDUCT.md

Awesome Lists containing this project

README

        
# _Common statistical tests are linear models_: Python port

[![Build Status](https://travis-ci.com/eigenfoo/tests-as-linear.svg?branch=master)](https://travis-ci.com/eigenfoo/tests-as-linear)

[![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/eigenfoo/tests-as-linear/master?filepath=tests-as-linear.ipynb)

A Python port of Jonas Kristoffer Lindeløv's post [_Common statistical tests are

linear models (or: how to teach

stats)_](https://lindeloev.github.io/tests-as-linear/), which originally had

accompanying code in R.

## Notes on Python port

The original post used R's built-in functions to verify that the linear models

were indeed equivalent to the statistical tests (by showing that the p-values,

t-values, and other such statistics, were the same in either case). In this

Python port, we instead verify that `scipy.stats` functions and `smf.ols` output

agree.

The original R post had [four

appendices](https://github.com/lindeloev/tests-as-linear/tree/master/simulations),

each of which demonstrated (through numerical simulation) that a common

statistical test was well-approximated by a linear model. These simulations have

not been ported to Python (yet!). This is [an outstanding

issue](https://github.com/eigenfoo/tests-as-linear/issues/14). In the meantime,

please refer to the [original appendices

upstream](https://github.com/lindeloev/tests-as-linear/tree/master/simulations)

for the simulations.

Finally, certain statistical functions and tests are supported by neither

`statsmodels` or `scipy`, the two statistical Python libraries used in this

port. R code with no simple Python equivalent (mainly in sections 6 and 7,

"Three or more means" and "Proportions: Chi-square is a log-linear model") has

not been ported. Such code has been noted in yellow warning boxes.

## Contributing

Please refer to [the contributing

guide](https://github.com/eigenfoo/tests-as-linear/blob/master/CONTRIBUTING.md)

for project structure information and development instructions.

## License

This work is licensed under a Creative Commons Attribution 4.0 International

License. Please refer to

[`LICENSE.txt`](https://github.com/eigenfoo/tests-as-linear/blob/master/LICENSE.txt)

for more details.