is there a way to construct a transition matrix of a time-homogeneous, finite Markov chain such that the stationary distribution always has full support (this is equivalent to all states of the chain being positive recurrent).

I am looking for something quick that doesn't involve drawing the state-transition graph and checking for absorbing states etc.

Best would be something on the level of "check that every column sums to one". Is this just wishful thinking?


When you write "the stationary distribution, I suppose you mean that the Markov chain is irreducible. An easy sufficient condition is that all entries are strictly positive (and that the sum of each row is $1$ so it is a transition matrix).

  • 1
    $\begingroup$ For a sparser pattern, tridiagonal with strictly positive entries on the sub- and superdiagonal is also fine. The main result to use is the Perron-Frobenius theorem. $\endgroup$ Jul 28 '14 at 6:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.