I updated to R v4.0.3 and updating my packages to run TMBhelper v1.3.0, and TMB v1.7.18. I am working with a non-converged model, but noticed that when I run TMBhelper::fit_tmb, the unconverged model returns NULL in the $opt spot.
I wanted to test if this was based on its non-convergence, so ran the following after a restart + workspace clear:
library(TMB)
runExample('randomregression') ## loads example obj into env
opt <-TMBhelper::fit_tmb(obj)$opt
which returned, after a few dozen mgc evals:
Error in sqrt(unlist(tmp)) :
non-numeric argument to mathematical function
As a workaround I am presently able to get $opt from nlminb() using my unconverged value, but wanted to flag this since it seems to be more than just my own model. I recognize that the above issues might be two separate problems, and the unconverged-model-fail might be a new feature. Thanks for your help.
I updated to R v4.0.3 and updating my packages to run TMBhelper v1.3.0, and TMB v1.7.18. I am working with a non-converged model, but noticed that when I run
TMBhelper::fit_tmb
, the unconverged model returns NULL in the$opt
spot.I wanted to test if this was based on its non-convergence, so ran the following after a restart + workspace clear:
which returned, after a few dozen mgc evals:
As a workaround I am presently able to get
$opt
fromnlminb()
using my unconverged value, but wanted to flag this since it seems to be more than just my own model. I recognize that the above issues might be two separate problems, and the unconverged-model-fail might be a new feature. Thanks for your help.