Dear @paul.buerkner, your example code is very close to my use case too. However, I’ve found that when applying get_prior()
onto the formula
for a problem similar to bform
, that there are no default priors given for estimating cor classes that would describe the correlations between beta1
, beta2
& beta3
, which seems standard in a hierarchical analysis when I read these teaching notes for a varying slope model. Does this mean that this non-linear hierarchical model isn’t 100% hierarchical because that correlation structure isn’t being modelled for? The default I’m used to in brms::
when doing something linear like y ~ 1+x1+x2 + (1+x1+x2|ID)
is that cor class priors are provided (even though this structure is implicit in the brm()
call) and they would also be estimated in the brmsfit
object.
Any ideas how I could force bmrs to add in that correlation structure for the non-linear hierarchical problem above?