States of Markov chain and stationary distribution

247 Views Asked by At

Let $X$ be a Markov chain with a state space $S={\{0,1,2,... \}}$ and a transition matrix $P$ with given $p_{i,0}=\frac{i}{i+1}$ and $p_{i,i+1}=\frac{1}{i+1}$, for $i=0,1,2,...$. Find out which states are transient, null and not-null. Find stationary distribution.

I have that $p_{0,0}=0, p_{1,0}=\frac{1}{2}, p_{2,0}=\frac{2}{3}, p_{3,0}=\frac{3}{4}$ and $p_{0,1}=1, p_{1,2}=\frac{1}{2}, p_{2,3}=\frac{1}{3}, p_{3,4}=\frac{1}{4}$.

I think that all states are persistent, so no states are transient.

There is a theorem which says that a state is null $\iff \lim_{n \to \infty} p_{ii}(n)=0$, so state $0$ is null. I guess all the other states are non-null but I don't know how to prove it. Also I don't know how to find stationary distribution.

Please help me with the above exercise, correct me where I'm wrong, and send any tips to the rest. Any will be much appreciated.

1

There are 1 best solutions below

0
On

You need to realise that since $p_{0,1} = 1$ and $p_{i,0} + p_{i,i+1} = 1$ then $p_{0,j}$ must be $0$ for all $j \ne 1$ and $p_{i,k}$ must be $0$ for all $k \ne 0 \mbox{ or } i+1$ .

Let $\pi_j = \frac{e^{-1}}{j!}$ for all $j \ge 0$. Then $\pi_j$ is the stationary distribution because

\begin{eqnarray} \sum_{i=0}^\infty \pi_i &=& 1 \ ,\\ \sum_{i=0}^\infty \pi_i p_{i,j} &=& \frac{1}{j} \pi_{j-1} = \pi_j \ \mbox{ for $j \ge 1$, and}\\ \sum_{i=0}^\infty \pi_i p_{i,0} &=& \sum_{i=0}^\infty \frac{i}{i+1} \pi_i \\ &=& \sum_{i=0}^\infty \left(1 - \frac{1}{(i+1)}\right)\pi_i\\ &=& e^{-1}\sum_{i=0}^\infty \left(\frac{1}{i!} - \frac{1}{(i+1)!}\right)\\ &=& \pi_0\ \ . \end{eqnarray} Consequently, all states are positive recurrent—none are transient or null recurrent.