Using conditional=True in predict (markov_autoregression)

[B Pontz]

Hi,

I was wondering if someone could help me understand what "conditional=True" does in predict for MarkovAutoregression. The docs say it makes predictions conditional on the current regime if it True. Usually the output from predict() is an array with a shape like (4939,)

 When you add conditional=True and using 2 regimes and an order of 1, it outputs a shape like this

(2, 4939)

What is the second array?

 I also noticed when you make conditional=3, it outputs a shape like this

(2, 2, 4939)

What's is the extra data?

Thanks,

Brian