I have a quite complex rstan model which I was able to run with 50000 iterations and 900 observations. My observations doubled and the model is not able to run anymore, it gets stuck at the initial phase of the estimation for days and no error message is shown. I have reduced the number of iterations to 25000 but still no luck in make the code progressing. Does anyone have a clue about what is going on?
Probably not related to why your model seems to be hanging, but how come you’re asking for so many iterations? The typical effective sample size should be close to the number of iterations for a well-behaving Stan model, and you should only need a couple thousand effective samples (total, across chains) to get good estimates of tail quantiles.
To get a better sense of what’s causing the hang, what interface are you using? Rstan? Pystan? Cmdstan?
Oops, you said rstan. Ok, try restarting sampling with the “refresh” argument set to 1 and see if you get more output.
2 Likes
Would using CmdStanR improve the speed? Rstan has been reinstalled on the server maybe there is some technical features that make it slow