Paper 3, Section I, H

Markov Chains
Part IB, 2014

Let (Xn:n0)\left(X_{n}: n \geqslant 0\right) be a homogeneous Markov chain with state space SS. For i,ji, j in SS let pi,j(n)p_{i, j}(n) denote the nn-step transition probability P(Xn=jX0=i)\mathbb{P}\left(X_{n}=j \mid X_{0}=i\right).

(i) Express the (m+n)(m+n)-step transition probability pi,j(m+n)p_{i, j}(m+n) in terms of the nn-step and mm-step transition probabilities.

(ii) Write iji \rightarrow j if there exists n0n \geqslant 0 such that pi,j(n)>0p_{i, j}(n)>0, and iji \leftrightarrow j if iji \rightarrow j and jij \rightarrow i. Prove that if iji \leftrightarrow j and iji \neq j then either both ii and jj are recurrent or both ii and jj are transient. [You may assume that a state ii is recurrent if and only if n=0pi,i(n)=\sum_{n=0}^{\infty} p_{i, i}(n)=\infty, and otherwise ii is transient.]

(iii) A Markov chain has state space {0,1,2,3}\{0,1,2,3\} and transition matrix

(1213016034014121200120012)\left(\begin{array}{cccc} \frac{1}{2} & \frac{1}{3} & 0 & \frac{1}{6} \\ 0 & \frac{3}{4} & 0 & \frac{1}{4} \\ \frac{1}{2} & \frac{1}{2} & 0 & 0 \\ \frac{1}{2} & 0 & 0 & \frac{1}{2} \end{array}\right)

For each state ii, determine whether ii is recurrent or transient. [Results from the course may be quoted without proof, provided they are clearly stated.]