`Eff.Sample` and `Rhat` for a combination of parameters

#1
  • Operating System: Mac
  • brms Version: 2.8.8

Let’s say I’m interested in the combination of two parameters, such a group contrast or an indirect effect. I then compute the parameter with the appropriate algebraic operation of two columns from posterior_samples(). Is there an easy way to compute the Eff.Sample and Rhat for such a combination?

#2

While waiting this to get CRAN relases, download https://github.com/avehtari/rhat_ess/blob/master/code/monitornew.R
and use rhat() and ess_mean() (and ess_quantile for quantiles, and mcse_* for mcse, etc.) for the derived quantities The input argument is “A 2D array without warmup samples (# iter * # chains).”

1 Like
#3

Very helpful. Thank you @avehtari!