Statsmodels: FAQ: Contributions Welcome

[bashtage]

Contributing

statsmodels welcomes contributions of all sizes, including:

docstring fixes, which can be very small

Documentation is hard, and many eyes help fix typos, missing parameters, and many other issues. We are always happy to accept docstring improvements. The doc builds on Travis CI contains a log of docstrings that aren't up to standards (about 6000 issues at the time of writing). Improvements to these are especially welcome.

notebooks

Examples are incredibly useful. If you have a written a notebook that highlights using sm in ways that are not already covered in examples, please share these to improve the documentation.

bug fixes

Bug fixes are always welcome. There are plenty of opportunities on the issue tracker to contribute.

enhancements

statsmodels is community-driven, and so enhancements are welcome. Large enhancements, for example, new models should:

• have clean, PEP8 compliant code;
• be tested against other reference implementations (R, SAS, Stata, etc.);
• be added to the appropriate section of the docs
• have a notebook that illustrates how to use the new model.

Smaller enhancements, for example, a new option on an existing method, are also welcome. More minor enhancements have a lower bar and should:

• have clean, PEP8 compliant code;
• be tested against other reference implementations (R, SAS, Stata, etc.) if possible, or checked for correctness if not (for example, if adding an option to select a subset of output from a function).

cleaning

Pure code clean-ups are generally discouraged. (Locally) cleaning up code as part of contributing to one of the other goals is encouraged.

[aidanrussell]

I'm trying to find the log of docstrings not up to standard you mention but I don't see it, can you clarify?

[bashtage]

If you look at the output of a docbuild run such as

https://travis-ci.org/github/statsmodels/statsmodels/jobs/732588884

and then search for

Checking Doc Strings

You will see a long list of docstrings that have missing variables or other issues.

[bashtage]

The most important types look like

statsmodels.discrete.conditional_models.ConditionalLogit.predict: Parameters {*args, exog, params, **kwargs} not documented

and

statsmodels.discrete.conditional_models.ConditionalLogit: Unknown parameters {groups}

since there are variables that are not defined or are most likely misnamed.