How to create a Mixture of First-order Markov Models

I have reprhased the question here: How to treat array observations in a custom loglikelihood