Running stan multiple times - best practice


I am trying to do a simple implementation for SBC in a specific model and I am wondering what the best practice is in terms of getting stan to run for e.g 1000 iterations via cmdStanR as an example. The Stan user guide does not seem to give advice on this.



I posted some code recently showing how to do things with the stantargets package.

More recently I’ve implemented full SBC using plain ol’ targets. Note that I have some modifications that I’ll be pushing later today that improve that code, so keep an eye out for those if you look at it. Note also that I’m positive that stantargets itself will soon be able to do full SBC itself too.


Also, @hyunji.moon is working on a package for SBC too and might have some input on the general topic of best practices for running cmdstan repeatedly? (@hyunji.moon: maybe check out what I cobbled together here; probably mostly redundant to what you’re doing, but possibly a source of ideas for alternative ways of doing things)


I will definitely check stantargets out.
And thank you for referencing your SBC code.