1

Let $(X_n)_{n\in\mathbb{N}_0}$ be a Markov chain with discrete state space $E$ and transition matrix $P$. Let $C\subseteq E$ be a communicating class. Prove or disprove the following statement. $C$ is periodic with period $p>1$ if $C$ can be decomposed into a disjoint union of sets $C_0\cup\ldots\cup C_{p-1}$ in such a way that if $i\in C_n$ and $j\in C_m$ with $m\neq n+1\text{ mod }p$, then $p_{ij}=0$. Distinguish between the cases when (a) $E$ is finite, (b) $E$ is countably infinite.

--

Do not see why it is necesary to distinguish between (a) $E$ finite and (b) $E$ countably infinite.

Anyway:

Assume that $p(C)=1$. Then $p(i)=1$ for a $i\in C$. Fix $i\in C$ and use it as a reference point for the disjoint sets $$ C_k:=\left\{j\in C| \exists~m\equiv k\text{ mod }p: p_{ij}^{(m)}>0\right\}. $$ These sets build a decomposition as described in the task. Because of this decomposition there exists exactly one $k\in\left\{0,1,...,p-1\right\}$ so that $p_{ii}^{(m)}>0$ for a $m\equiv k\text{ mod }p$.

Moreover it exists a $\tilde{n}\in\mathbb{N}$ so that $p_{ii}^{(n)}>0$ forall $n\geqslant\tilde{n}$, i.e. $$ S:=\left\{\tilde{n},\tilde{n}+1,\tilde{n}+2,\ldots\right\}\subseteq R(i):=\left\{n>0|p_{ii}^{(n)}>0\right\}. $$ Write $S$ as $$ S=\bigcup_{k=0}^{p-1}\left\{\ell\in S|\ell\equiv k\text{ mod }p\right\}. $$ Then $p_{ii}^{(s)}>0$ for all $s\in S$ implies that $i\in C_t$ for all $t\in\left\{0,1,...,p-1\right\}\setminus\left\{k\right\}$ contradicting that the sets $C_j, j\in\left\{0,1,...,p-1\right\}$ are disjoint.

So $p(C)>1$.

--- Again: Why is it necessary to distinguish between $E$ finite and $E$ countably infinite?

mathfemi
  • 2,631
  • You know that all this is explained in detail in tons of online resources, do you? – Did Nov 01 '14 at 22:10
  • You might want to check Norris, Markov chains. (ed ajf) – Did Nov 01 '14 at 23:10
  • @Did I have this book but where is it in this book? I do not find it! – mathfemi Nov 01 '14 at 23:12
  • Try pages 18-20 there. – Did Nov 02 '14 at 01:08
  • @Did Maybe you can say a word or two to my last edit? – mathfemi Nov 03 '14 at 11:25
  • What a thorough cleansing of the comments... OP: Did you ask for it? That the request was fulfilled was probably at least partly ill-advised because this thread of comments contained an analytical description of the problem with this OP's recent string of questions, equivalent to asking for the content of the first chapters of a book on the classical theory of Markov chains. Is this considered an undesirable message on math.SE? – Did Nov 03 '14 at 12:24
  • I did not ask to delete the comments. 2. What about my edit/ proof?
  • – mathfemi Nov 03 '14 at 15:04
  • @Did I do not know exactly why you do not want to say a word to my edited proof.. I think it might be correct and therefore I would like to get a feedback to my efforts. Moreover I would like to know wh it is necessary to distinguish between $E$ finite and $E$ countably infinite, because the proof that I gave holds for both cases as far as I am able to see. – mathfemi Nov 03 '14 at 21:15