This would make it possible to obtain Empirical Bayes posterior predictive distributions using simulation. The user could supply a function for the derived quantity of interest. Eventually, it would be nice to account for uncertainty in the hyperparameters, but that would be harder.
This would make it possible to obtain Empirical Bayes posterior predictive distributions using simulation. The user could supply a function for the derived quantity of interest. Eventually, it would be nice to account for uncertainty in the hyperparameters, but that would be harder.