When we test the analytical solutions in stan against their R equivalents we sometimes see NANs in the stan output that are 1 in the R output. This indicates some kind of overflow or boundary issue to me.
So far we have seen no evidence of this being an issue when fitting models indicating it might be related to our testing setup pushing towards extreme edge cases but regardless this needs investigation.
Originally posted by @seabbs in https://github.com/epinowcast/primarycensoreddist/issues/109#issuecomment-2402157878
When we test the analytical solutions in stan against their R equivalents we sometimes see NANs in the stan output that are 1 in the R output. This indicates some kind of overflow or boundary issue to me.
https://github.com/epinowcast/primarycensored/blob/main/tests/testthat/test-stan-primarycensored_analytical_cdf.R
So far we have seen no evidence of this being an issue when fitting models indicating it might be related to our testing setup pushing towards extreme edge cases but regardless this needs investigation.