Rstanarm/brms/lme4 grouping term question

If I had multiple groupings of data (like age and state for a bunch of people polled), is there an advantage to saying:

(1 | age) + (1 | state) + (1 | state:age)


(1 | age) + (1 | state:age)

Seems like you’d want the first, right? The second is making it harder for state level effects?

I think the first would yield two state effects in rstanarm or brms and not sample well. But the two would be essentially the same for lme4 because it integrates all that out anyway.

1 Like