Log lik for new group data in hierarchical multinomial model

Modeling
,
2

Does this thread help? It feels like a similar problem where the solution is to use _rng() functions for the unobserved groups.