Bayesian Markov Switching Models

I am working on a rstanarm-esque implementation of Markov Switching Models with/out time-varying transition probabilities for time-series and panel data. I’ve a basic working version for the non-time varying version done (though probably riddled with errors) but was interested in what people who may be likely users would be interested in in terms of functionality. Right now it works with shared/individual transition probabilities and shared and individual state sequences and user-defined constraints to prevent label switching for arbitrary numbers of states. The TVTP version will allow for the same things as well as general, state specific and state-state specific predictors for the TVTP.

I’d be particularly interested in any kind suggestions for ppc that could be implemented but any ideas or requests are welcome.