strengejacke: That package sounds great, but doesn’t actually seem to work with brms multinomial models:
> ggpredict(mod, terms = "x1", ci.lvl = 0.95, type = "re")
Note: uncertainty of error terms are not taken into account. You may want to use `rstantools::posterior_predict()`. Error in `$<-.data.frame`(`*tmp*`, "predicted", value = c(0.249904418608922, : replacement has 8 rows, data has 2
Will likely be useful once the bugs are fixed though.