ENH: stats.fisher_exact: convert output tuple to Bunch

[JozsefKutas]

Reference issue

#16364
#13118

What does this implement/fix?

#16364 suggests that all statistical functions that currently return tuples should return Bunch objects. This PR converts the tuple returned by stats.fisher_exact to a Bunch.

[mdhaber]

Thanks @JozsefKutas. At first glance, this looks great! Let's see what another maintainer thinks of gh-16387, and if that looks OK to them, I'll review/merge this and other follow-ups.

[tupui]

I would prefer to have a common "tuple bunch" as I said in the other PR from @mdhaber. Otherwise I am +1 on doing this.

[mdhaber]

Ok, what should the name be? StatsTestResult?

[tupui]

Might be too general. What about HypothesisTestResult? Not sure if there would be something that would just say p-value and statistic. SignificanceTestResult maybe?

[mdhaber]

Is combine_pvalues well-described by "hypothesis test"? Not sure, but I like a slight tweak of what you wrote: I think SignificanceResult is a good name for an object with just attributes statistic and pvalue.

[tupui]

Let's go with that then 👍

[mdhaber]

OK I just changed gh-16387 to SignificanceResult; please merge that if it looks good. Then we can update this and continue. Let's leave SignificanceResult out of the public API for now, but at some point we may wish to expose it.

[mdhaber]

Merged main so we could use SignificanceResult and parametrized test.

@JozsefKutas do these changes look good to you? If so, I'll merge, and we can continue down the list - just like this is great, I think.

[JozsefKutas]

Yup, that seems fine to me.

For the other statistical functions, both chi2_contingency and median_test return other results as well as the statistic/p-value. In these cases would you prefer a function-specific result object, as per the original PR?

[mdhaber]

would you prefer a function-specific result object, as per the original PR?

Yup, I think so. We can consider consolidating similar result objects later.

[mdhaber]

Thanks @JozsefKutas @tupui!