No, because those don’t (usually) have a “closed form” expression with which to generate random draws directly. You can always use NUTS to do prior predictive sampling by just omitting the “observed” from the model.
2 Likes