Data, parameters and model blocks around the generated likelihood

prior_uniform() and prior_normal() are written out through the data block, so one compiled model serves any bounds and any prior mean. Every other named family (see prior_families) becomes a literal Stan sampling statement instead: its parameters are what pick the distribution, and there is nothing left to vary at run time. A prior built with prior_custom() is arbitrary R code with no Stan counterpart, so the model block is the user's to write.

Usage

stan_model_blocks(fit, name, packed)