I searched but did not find any way to automatically incorporate evaluating the log posterior probabilities of a test set in brms, as described here. Is this possible, either now or as a feature in the future?
Check out brms::log_lik.brmsfit, which accepts a newdata argument.
Excellent thanks!