make chain-id accessible to Stan program?

[Sebastian Weber]

Hi!

I just had a case where I urgently needed the chain id in the transformed data block in order to select a subset of my input data.

This could be useful for a bigger audience, i.e. this could be used to conveniently prepare a single data set and then do a cross-validation with Stan. If the chain id would be available in Stan as a variable, then the user can select the subset to leave out based on the chain id.

Does that make sense?

If yes, I am happy to file an issue for that on the stan repo.

Best,
Sebastian

PS: I solved my problem by copying my data-set around and pasting into it the respective chain id. Not a nice solution as I find.

[Bob Carpenter]

Feel free to file an issue with concrete requests.

It looks like the notion of data is changing from
my original conception, where data was fixed once
and then shared among all chains.

With (P)RNGs in transformed data, it's no longer fixed.

If we do this, then it might vary by chain? I'd rather
not mix the concept of chain and subset of data, because
if chains have different data, we won't be able to run
things like convergence diagnostics.

We should discuss where data's going at a meeting. I'll
hopefully be on next week's call.

- Bob

> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.