Mixed effects for multivariate rescor in brms?

Interfaces brms
1

Is it currently possible to estimate heterogeneity in residual correlations by groups using mixed effects with brms?

I’m interested in estimating a multivariate individual-participant data meta-analysis using brms to examine variability in correlation matrices among a set of variables across samples.

For example, I’d like to be able to estimate random effects for the residual correlation parameters for the following model.

library(brms)
d <- psych::bfi |> 
  transform(education = ifelse(is.na(education), "NA", education)) |> 
  subset(select = -c(gender, age)) |> 
  na.omit()
m <- brm(
  bf(mvbind(A1, A2, A3, A4, A5) ~ (1 | education)) + 
    set_rescor(rescor = TRUE), 
  data = d, 
  cores = 8, 
  threads = 2
  )
1 Like
2

Quick response: My bet is not yet. There’s a GitHub issue on random effects for residual correlations from a while back and I’m not aware that it has been resolved.