Closed jamessteeleii closed 2 years ago
I've just realised the error above was due to me having fit the models in parallel. When I set pp_check(model, cores = 1)
the above error doesn't occur, but I do get the following:
Error in rbeta(n = N, plogis(mu) * phi, (1 - plogis(mu)) * phi) :
invalid arguments
The same errors occur when I try to reproduce the model from the vignette and then perform pp_check()
too. I can't tell if it's something wrong locally for me or not though. Of course, this applies to other similar functions for posterior predictive draws too.
Oh, the exception seems to be insight::get_predicted
. That seems to work.
I can't reproduce this error. The vignette runs fine and uses pp_check
. Maybe you should use that dataset and see if it works on your system as well.
Thanks Robert - yep I tried to just run the vignette example using the pew
dataset, but same errors. I got a friend to try and it worked fine for him so must be something local to my system causing issues. I tried to uninstall the package and reinstall but same issue. It's not the end of the world as insight::get_predicted
works for some reason to get the posterior predicted draws. I may try to completely uninstall everything and reinstall at some point and see if that resolves it.
Yes the vignette works fine with the current version of idealstan, so without some way to replicate this I'm going to need to close the issue. However, if you can help me reproduce then I can try and fix whatever is broken.
I've noticed the few times I have fit ordbetareg models that, while I can use plot() to produce trace plots to inspect the model, pp_check() doesn't appear to work. I get the following error each time I try to use the function on an ordbetareg model: