Albert Steppi

> @steppi perhaps this is the catastrophic cancellation you were looking for? Sorry, when I reviewed gh-16605 I'd assumed that `log_diff` was essentially using the trick from my comment [here](https://github.com/scipy/scipy/issues/13923#issuecomment-825118292)....

> I'll continue to look into the theory, but @steppi if you know or are interested in the theory of the bootstrap, this is (yet) another one I could use...

> If you could help check my work - perhaps by implementing your own calculation of the equation 15.36 - I'd really appreciate it! Sure! I should have some time...

@mdhaber, hey I haven't forgotten about this, just in a very busy period now. I hope to look into it over the weekend. Please ping me on anything else that...

> Thanks so much for looking into this! I'll add the mouse data acceleration parameter example and the comparison against R as a test, then I'll be done with this....

@mdhaber Oh sorry for merging before your tests finished. Just getting to this now. I saw that the only docs you touched built correctly so figured it was OK to...

> @steppi if you also like linear algebra, this may be interesting to you. Putting this next in my queue.

The failures on 32 bit is because it compares `wright_bessel(a, b, 0)` against `rgamma(b)`, and `wright_bessel` no longer uses the C implementation from `scipy/special/cephes` for this case, but instead the...

> (1) it looks like `_evalpoly.pxd` is removed, did this already have a corresponding `_evalpoly.h` file in a different migration PR? Yes, this was removed in the PR translating the...

> CI failure looks real Interesting. It looks like that job was added only a few weeks ago. I'll have to look into it.