R package "brms" model specification

Modeling
, ,
1

I’m relatively new to Bayesian Regression Models and currently exploring the R package “brms” (rstan). My current focus is on developing a regression model for a multilevel dataset with hierarchical groupings (var4/var3/var1) and repeated measures over various time occasions (var2). To accommodate temporal autocorrelation, I aim to specify the autoregressive-1 covariance structure. Could someone assist me in correctly specifying the model’s random effects?

Here’s my current model setup:
bf(
mvbind(var9, var10, var15, var16) ~ var2 + var13 + var14 + var19 + var20 + var34 + var46 + var47 + var50 + var51 + var52 + var56 + var57 + var59 + var60 + var62 + var65 + var66 + var89 + var90 + var91 + var92 + var93 + var94,
zi ~ var2 + var5 + var114,
autocor = ~ar(time = var2, gr = var4:var3:var1)
)

However, I’m uncertain whether this specification will yield the desired result.

Furthermore, I’m seeking advice on setting priors to enhance the accuracy of the estimates. Any guidance on this aspect would be greatly appreciated. Thank you!

2

Hey @bhara033, it looks like this is your first post. Welcome to the Stan forums.

It’s hard to follow along with such a complicated model. Consider presenting a simplified model that only has the bare essentials for your question. You’re more likely to get engagement that way.

1 Like
3

Thank you for responding! I will repost a simplified version of the problem.

1 Like