0
$\begingroup$

How do we prove that an irreducible, doubly stochastic (row and column sums to 1) discrete time markov chain with infinite state space cannot be positive recurrent?

Initial Ideas: I think the place to start is naturally trying to proving that the stationary distribution cannot possible exist for this setup...but I am having trouble getting started with proving this.

Thanks!

$\endgroup$
1
$\begingroup$

Let $P$ be the transition kernel and let $\nu= \mathbf 1$. Then for each $k$ we have $$(\nu P)_k = \sum_j \nu_jP_{jk} = 1, $$ so $\nu$ is an invariant measure for $P$. Since $P$ is irreducible, any stationary distribution for $P$ must be a multiple of $\nu$. But $\sum_k \nu_k=\infty$, so there cannot be such a distribution.

$\endgroup$
  • $\begingroup$ Did you mean to say can't? $\endgroup$ – Cehhΐro Oct 14 '16 at 6:58
  • $\begingroup$ Yes, thank you. $\endgroup$ – Math1000 Oct 14 '16 at 8:09
  • $\begingroup$ Thanks!! Just a follow up..the fact that any stationary distribution is multiple of invariant measure encompasses periodic markov chains I assume? It is just a result I am not familiar with in general.. $\endgroup$ – user1421195 Oct 14 '16 at 18:52

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.