Added module for ztest and testing as well

[cassielayden]

Reference issue

Progress made on feature request #13662

What does this implement/fix?

Implementation adds a module for performing the z tests and a test suite for the Z-test module as well.
Z-Test module is at scipy/stats/ztest.py
Test module is at scipy/stats/tests/test_ztest.py

Additional information

The implementation is not complete. The functions in the added module are correct and working when provided the input outlines. Comments are included throughout the code that explain how and where input should be taken in a user interface. Feel free to contact me with more questions about this.
Additionally, I will attempt to implement the interface as I see possible, however I do believe another developer may be better fit for this task.

[tupui]

Thanks @cassielayden for submitting a PR. I am afraid I will have to close it for a few reasons.

From the issue this PR is addressing, @charlotte12l already expressed some interest in working on this. While there is no formal assignment of issues/PR, we are all volunteers and try not to re-do the same work. We certainly don't want to have contributors competing on PRs. Instead we prefer to share the work load by discussing in issues or on the mail list before.

Thus, I would invite you to contact @charlotte12l and see how you can be of help if this topic interests you. In the present case, a new statistical test, we also would want to design if in a way we could reuse existing components from other existing tests. In terms of maintainability, we would not want to include an implementation which would be totally contained here.

Before working on something else, I would also invite you to read our contributing documentation. It's a long document, but it's really important that you understand it and follow it. There are valuable informations such as how to format your code and tests. In the end we will have to respect this document from start to finish before considering including any contribution. As for other issues you could have a look at, we do have a label good first issue.

Thank you for your understanding and your willingness to contribute to SciPy. While this PR was not successful, I do hope that you would consider contributing again.

[ilayn]

It's better to first check if @charlotte12l is actually working on it. Looks like it is stalled over there so it doesn't have to be closed in case they don't ping back.

[rgommers]

Thanks for reopening @ilayn. @tupui this PR needs some work but it's totally valid to work on this. I suggest only closing a PR if you're really certain that's the right thing to do (like, it stalled for a year and someone else took it over in a new PR, or the bug it addressed was already fixed.

[rgommers]

A few high-level comments to get started.

I suggest first writing the public function, something like

def ztest(...):
    """
    Docstring here
    """
    ...

[rgommers]

This is a new file, so it does not need to have this copyright copied over from another file. Please remove this block comment.

[rgommers]

Please import only the functions and modules you are actually using to implement the Z test functionality. Running flake8 on this file will show you unnecessary imports.

[rgommers]

Same here, just add __all__ = ['ztest'] (if that's the function name).

[rgommers]

Can you add a comment about where these values come from?

Also, is 4 digits precise enough, and is there not a way to calculate these values instead?

[bashtage]

These are the cdf of a normal. Why not use stats.norm.cdf?

[charlotte12l]

@tupui @ilayn @rgommers Sorry for the late reply. I'm currently not working on it because there is not much interest after I mailed the mailinglist about this, so I added this to the end of my GSoC timeline.
@cassielayden I read your codes and I'd recommend taking a look at gh-12873 for a good example of implementing a new hypothesis test(Thanks @mdhaber for mentioning the example in #13662 ).
Moreover, maybe it is better to return the p-values rather than 0/1 to make it consistent with other tests? You can use _normtest_finish to get the p-value from the Z statistic.
Finally, z-test is quite similar to t-test, I think you can always refer to t-test implementation for naming, code structures, tests, logics, etc. :)

[tupui]

@tupui @ilayn @rgommers Sorry for the late reply. I'm currently not working on it because there is not much interest after I mailed the mailinglist about this, so I added this to the end of my GSoC timeline.

No worries, and thanks for helping with this specific information 😃