Add split normal distribution

[keatonb]

Split normal is a useful distribution that isn't currently implemented in SciPy. I will work to implement.

Split normal: https://en.wikipedia.org/wiki/Split_normal_distribution

[tupui]

Hi, thanks for the proposal. Usually we use first the mail-list to introduce new features like a new distribution. I invite you to post your proposal here with sound motivations (like strong literature indicating the value of the feature): https://mail.python.org/pipermail/scipy-dev/. This avoids starting to work on something and having the risk of the PR being rejected.

[keatonb]

Good to know. I'll do that. The instructions on proposing a new distribution doesn't mention the mailing list, fyi, just says to open a GitHub issue: https://docs.scipy.org/doc/scipy/reference/dev/contributor/adding_new.html

[tupui]

Thanks for pointing it out. We are always looking at unifying and improving the dev docs. I will mention this.

[mdhaber]

@keatonb if mailing list response is positive, a good example PR for a statistical distribution - and one that may be particularly relevant because it was also for a piecewise-defined distribution - is gh-12694.

[czgdp1807]

@keatonb @tupui Is this issue approved on the mailing list? If not then I can try proposing it. Please let me know.

P.S. If it's already approved then please share the link to the thread. It would be great to go through it before starting working on it.

[tupui]

@czgdp1807 not sure I saw such a message. You can go through the archive here https://mail.python.org/pipermail/scipy-dev/

[keatonb]

@czgdp1807 sorry I didn't get around to it. Feel free to propose for this on the mailing list.

[cmdupuis3]

Taking...

[tupui]

@cmdupuis3 do you mean that you are going to ask the mailing list? This is a requirement before starting any work as you can read from the previous comments.

[cmdupuis3]

I guess so? Someone told me to try working on this at the SciPy hackathon as a good next-step issue.

Looks like the links here are out of date. Do I just start a thread on the new mailing list?

[tupui]

Yes sure feel free to send an email to the list 😃

[mdhaber]

Someone told me to try working on this at the SciPy hackathon as a good next-step issue.

Yes, that was me.

Yes, it's best to start with a post on the mailing list before starting work because ultimately it will need approval there before being merged. Better to know sooner than later.

Please see (here)[https://docs.scipy.org/doc/scipy/dev/contributor/adding_new.html#adding-a-new-statistics-distribution] for more information about adding a statistical distribution. You can also search the (closed) PRs for "asymmetric Laplace". The PR adding that is a good, recent example.