Chain mixing order in brms log_lik output

Interfaces brms
1

I’m using brms’ log_lik function to extract the log-likelihood from a brmsfit object. I would like to use relative_eff from rstan, but I’m not sure in what order log_lik returns the different chains’ log-likelihoods. Does it just append the log-likelihoods from each chain? What I mean is, if I have a toy brmsfit object, “fit”, with two chains, “a” and “b”, and three samples, does log_lik(fit) return the values in the order c(a,a,a,b,b,b), or does it return c(a,b,a,b,a,b).

Thanks for your help!

2

Sorry this question fell through the cracks and no one responded yet.

I think log_lik is stacking the chains on top of each other, so it will put all the iterations from chain 1 first, then chain 2 below those, then chain 3, etc.