The newest version of glmmTMB is about to go to CRAN. We changed the parameterization of the Gaussian dispersion from log(variance) to log(SD), which changes results slightly and made one of your comparisons testing with glmmTMB fail (the deviations were only 3e-08 on average, but this is larger than the default expect_equal() tolerance of ~ 1.5e-08 ...) This PR loosens the tolerance for that test slightly (to 1e-05; 1e-06 or 1e-07 would probably be sufficient but I don't think it hurts to leave it at 1e-05). Since this is a tiny change you could always reject the PR and change the tolerance setting however you preferred ...
seananderson
commented
6 months ago
The newest version of
glmmTMBis about to go to CRAN. We changed the parameterization of the Gaussian dispersion from log(variance) to log(SD), which changes results slightly and made one of your comparisons testing withglmmTMBfail (the deviations were only 3e-08 on average, but this is larger than the defaultexpect_equal()tolerance of ~ 1.5e-08 ...) This PR loosens the tolerance for that test slightly (to 1e-05; 1e-06 or 1e-07 would probably be sufficient but I don't think it hurts to leave it at 1e-05). Since this is a tiny change you could always reject the PR and change the tolerance setting however you preferred ...