Sean Pinkney
Sean Pinkney
No worries, enjoy your vacation!
@rok-cesnovar can you restart the CI/jenkins process? Once that finishes I can merge in
@rok-cesnovar (or @bob-carpenter or @bgoodri) I updated this to use the `ldlt` decom rather than the `cholesky` as it's more numerically stable. I attempted to keep the `ldlt` factor type...
> Taking the square root is where most of the precision gained from using LDLT gets lost. There should be a way to differentiate with respect to the diagonal elements...
The issue is that the cholesky decomposition is less stable than the ldlt so I'm waiting until tuples get in the language or once someone implements ldlt with derivatives.
A few notes. It is possible to run all the tests for a particular function if you have `stanc3` built. Let's first build `stanc3` (tested on a mac) - Add...
@bob-carpenter this PR to my helpful stan functions repo has the code and a reproducible R example. Same as attached by Ethan but possibly a bit cleaner. https://github.com/spinkney/helpful_stan_functions/pull/13/commits/e603fb9c815186ce628cd441371c5270877afe55
To @betanalpha's point this is pretty straightforward. Here's the overloaded binomial and poisson cdfs in R ```r > library(gsl) > cpoisson_cdf > cpoisson_cdf(2, 5) [1] 0.124652 > ppois(2, 5) [1]...
@bob-carpenter can't we get around that by defining new functions for the continuous/relaxed version of discrete cdfs? So we could have a "relaxed_poisson_cdf" or "continuous_approx_poisson_cdf" or whatever name people like...
It's not "wrong" it's the "continuous" or "interpolated" version that doesn't use floor...