This is intended to make work on sample posterior predictive easier by making the arguments it takes more predictable.