fancidev

> This is something I've considered adding to similar functions before. @tirthasheshpatel and I will be working on other aspects of `trim_mean` shortly and we will consider this again at...

It would help those who build scipy on M1 Mac (like me). For me the build was smooth other than some linker warnings (about unsupported switches) which I safely ignored....

Maybe worth coding an explicit CDF and PPF using formulas from Wikipedia.

> Are there difficulties for very high moments for the powerlaw distribution? No, the `expect` function works perfectly fine for 5th moment and only starts to lose precision when moment...

@dschmitz89 Could you please merge and close this PR?

> @fancidev I do not have commit rights, so cannot merge. You could not know that though, that's fine :). Still, I would rather not discuss scipy's ACL openly here....

I encountered the same issue when looking at the `vonmises_fisher` distribution: `ive(0.5, 1e16)` returns `nan`. For the specific case of `argus`, a workaround is to replace `ive(1, x)` with `i1e(x)`,...

A possible trick is to use `xp.max` instead of `xp.sum`. Though for readability `any(isnan(x))` is clearly best. And since premature optimization is the root of all evil, it may make...

In addition, `numpy.sum` performs pairwise summation, which is an overkill for detecting `nan` … It would be best if the Array API had a `count` function, so that you could...

Is there a place to see the current progress of this branch of development?