I'm modeling markov chains for some data sets.But not all of them has unique stationary state.
I create the transaction matrix A from the data sets. If Ahas no unique stationary state, How can I change A to get a proper approximate unique stationary state for my markov chain modeled from the data set?