Christian Lorentzen

The R package [expectreg](https://cran.r-project.org/package=expectreg) might be a good reference for unit tests.

I might need some help to make it work with `axis` and `nan_policy`.

@mdhaber I think I responded to and addressed all review comments such that another round of review would be very welcome.

Mailing list: https://mail.python.org/archives/list/[email protected]/thread/BFLKYOSA3CMEHOEPZ56E5IPWPSMEN2BF/

Can someone confirm that the email made it through the spam filter?

@mdhaber I don‘t mind at all. Please go ahead. Your help is highly appreciated.

@mdhaber Based on your first 2 point2, I expanded the Notes section quite a bit, see https://github.com/scipy/scipy/pull/17039/commits/f9e11b76af6f321fa551c80ab6a0ef4585d1a7ae.

About the minimization property: The 1rst order condition $E_{Y\sim F}[V(z, Y)]=0$ with an identification function $V(z, y)$ is, in a way, unique. (If $V(z, y)$ is an identification for the...

@mdhaber @h-vetinari Thanks for the reviews. It was a pleasure to interact with you.

I think a good starting point would be to have available a simple empirical expectile, such as `np.mean(x)`, without the *regression* stuff. I thought, scipy would be a good place...