0

Suppose that you have a Markov chain with state space $E$ containing $0$. Assume that $$ p_{00}^{(2n)}=\binom{2n}{n}\left(\frac{1}{2}\right)^{2n-1}~~~\text{ and }~~~p_{00}^{(2n-1)}=0~~~\text{ for }n\in\mathbb{N}. $$ Is $0$ transient, positive recurrent or null recurrent?

Already showed that $0$ is recurrent.

Claim: $0$ is null recurrent.

Therefore have to show $E_0(t(0))=\infty$, where $t(0)$ is the first return time of $0$.

Question: Is it right that $$ P_0(t(0)=2k)=\frac{1}{k}\binom{2(k-1)}{k-1}\left(\frac{1}{2}\right)^{2k-1}? $$

I continued with the following:

\begin{align} \mathbb{E}_0(t(0))&=\sum_{k\geq 1}2k\cdot\underbrace{\mathbb{P}_0(t(0)=2k)}_{=\frac{1}{k}\binom{2(k-1)}{k-1}\left(\frac{1}{2}\right)^{2k-1}}\\ &=1+2\sum_{k\geq 2}\binom{2(k-1)}{k-1}\left(\frac{1}{2}\right)^{2k-1}. \end{align}

Using Stirling's formula $n!\sim\sqrt{2\pi n}\left(\frac{n}{e}\right)^n$ I get \begin{equation*} 1+2\sum_{k\geq 2}\frac{1}{\sqrt{\pi (k-1)}}\cdot 2^{2(k-1)}\cdot \frac{2}{2^{2k}}=1+\sum_{k\geq 2}\frac{1}{\sqrt{\pi (k-1)}}=1+\sum_{k\geq 1}\frac{1}{\sqrt{\pi k}}\geq 1+\frac{1}{\pi}\sum_{k\geq 1}\frac{1}{k}=\infty. \end{equation*} Thus $\mathbb{E}_0(t(0))=\infty$. That is $0$ is null recurrent.

mathfemi
  • 2,631
  • As has already been mentioned to you (rather exhaustingly), (1.) your formula for $p_{00}^{(2n)}$ is bogus and (2.) there is no need to compute $P_0(t(0)=2k)$ to determine null/positive recurrence (as explained here). Regarding the question explicitely stated here, namely, to confirm whether your formula for $P_0(t(0)=2k)$ is correct or not, note that you produce zero motivation/explanation for this formula. .../... – Did Nov 16 '14 at 10:52
  • .../... Explaining how you reached it is mandatory (and nearly everything else in the current question is actually offtopic). Until then, this question should be closed for "lack of context". – Did Nov 16 '14 at 10:52
  • @Did I do not know what you mean with that my formula for $p_{00}^{(2n)}$ is bogus, which formuly do you mean? – mathfemi Nov 16 '14 at 10:56
  • ?? The only formula starting with $p_{00}^{(2n)}=...$ in this post. // Listen, you will not drag me into another morass of a convoluted discussion where you basically never answer what I say. Hence, this is my last comment here. – Did Nov 16 '14 at 10:59
  • Okay, but this formula is given in the task is not my result. – mathfemi Nov 16 '14 at 11:01
  • Ok, I try to explain how I got the formula for $P_0(t(0)=2k)$. I looked how many ways there are starting in 1 going back to 1 in 2k-2 steps (2 steps are reserved for going from 0 to 0 and then in the end going back from 1 to 0) and which probability each such way has. – mathfemi Nov 16 '14 at 11:17

0 Answers0