DOC: Rename `*args` param in `f_oneway` to `*samples`

The current signature for the f_oneway function in the stats module reads:

def f_oneway(*args, axis=0):

which does not provide any real clarity, imo.

My suggestion is to refactor the *args param to *samples, or *samps—whatever form is more agreeable. The docstring itself uses sample:

sample1, sample2, … : array_like

Of course anyone calling this function would have an idea of what the function would accept; nevertheless, I think this would add value at zero cost (just a refactor within the function).

[tupui]

Sure—happy to 👍 Should I do the other functions as well? @kruskal, levene, bartlett, fligner

I would say so.

Regarding the docstrings: couldn’t find anything wrong with them, but maybe I’m overlooking something?

sample1, sample2, … : array_like *args : tuple of array_like

These are the two formats I found; do we prefer the second form, or the first (*when the param is a arg)

I would go with the second form *samples : tuple of array_like. It’s what numpydoc says.

Had one question on the tests before submitting the PR—I’ve already run the entire test suite once today; after the change, is it sufficient to run python[3] runtests.py -v --submodule stats? Or is it required to rerun the entire suite before pushing?

Thanks for asking! The more you test locally the better as you could limit CI usage. Still, it’s usually fine to just run the tests of the module you are touching. Of course, some functions are used in other modules and it might be disruptive. But it’s ok and the CI will catch this. Something that is more important I would say would be to skip some CI if you just change the doc for instance (cf CI doc on our dev doc).

[tupui]

Thanks for reporting @ShivnarenSrinivasan! Seems like a good idea and the docstrings would also need an update. samples seems more appropriate. We should avoid truncating words like this. Would you like to make a PR?