Sean Pinkney
Sean Pinkney
This is an interesting model. Can you make an argument for this being in the docs vs a case study?
> For two change points, that's O(T^3) and O(T^2). Yep! Also, we need to fix that the matrix `log_unif` prior needs to be constructed as `rep_matrix(log_unif, T, T)` not `rep_matrix(log_unif,...
@bob-carpenter do you have the dynamic version for 2+ change pts? If you put it here I can make a pr with the relevant changes.
I may remove the current 2+ cut point comment and add a variant of your description for the 2+ case without actually showing it and mentioning that this is beyond...
btw a vectorized `log_Phi()` that is robust (as in your posted linked issue) would be great to have for copulas
Part of this is a math issue though. Since `normal_lcdf(x | 0, 1)` does not output the same as `Phi()` the tails. Issue: https://github.com/stan-dev/math/issues/2470#issue-856948197
For functions * I believe all reverse mode analytical derivatives are in https://github.com/stan-dev/math/tree/develop/stan/math/rev/fun * Similarly for forward mode ../fwd/fun For distributions this is a bit more difficult. Though if `partials`...
> Is it safe to assume that if one signature of a function has analytic [fwd|rev] derivatives that the vectorized versions/overloads will? I believe so. @andrjohns has done a lot...
Thank you!
@andrjohns, I don't see a rev version of `hypergeometric_3F2` so we probably shouldn't expose that now. I also see that `hypergeometric_pFq` has a comment about not being exposed to users...