Combining posteriors


first of all, thanks for all the great work you’re doing!

I’ve been using brm_multiple before and I was hoping rstanarm had something like that, i.e., to use multiple datasets (with imputed values) and combine them all in one posterior. I have looked and looked but it doesn’t seem to be the case.

So, my question is how one could handle this in rstanarm? I looked through the brms source code it seems Paul does some magic and then combines the posteriors somehow. I see two approaches in rstanarm that might work and I’d be grateful for any input:

  1. Somehow just bunch all posteriors together into one? Is there an easy way to do this (combine_models in brms seem to do this for brmsfit objects)?
  2. Create an MLM where datasets is one level?

There isn’t anything like (1) in rstanarm but you could try doing (2). For some purposes, you don’t need the stanreg object and can just make a list of as.matrix(). If you do need the stanreg object for posterior predictions or something, then you can make a list of those matrices.

1 Like