fix issue #909

R/conditional_effects.R

@@ -586,7 +586,7 @@ get_int_vars.mvbrmsterms <- function(x, ...) {
 
 #' @export
 get_int_vars.brmsterms <- function(x, ...) {
-  advars <- ulapply(rmNULL(x$adforms[c("trials", "thres")]), all_vars)
+  advars <- ulapply(rmNULL(x$adforms[c("trials", "thres", "vint")]), all_vars)
   unique(c(advars, get_sp_vars(x, "mo")))
 }
 

doc/brms_customfamilies.Rmd

@@ -237,7 +237,7 @@ For the purpose of the present vignette, we will go with the latter approach
 expose_functions(fit2, vectorize = TRUE)
 ```
 
-and define the required `log_lik` functions with a few lines of code.
+and define the required `log_lik` functions with a few lines of code[^phi].
 
 ```{r log_lik}
 log_lik_beta_binomial2 <- function(i, prep) {
@@ -249,6 +249,9 @@ log_lik_beta_binomial2 <- function(i, prep) {
 }
 ```
 
+[^phi]: The presented post-processing functions need to be adjusted if you
+predict `phi` in your model as well by writing `phi <- prep$dpars$phi[, i]`.
+
 With that being done, all of the post-processing methods requiring `log_lik`
 will work as well. For instance, model comparison can simply be performed via