Stochastic

# Case 2 now suppose that z x reasoning as above we

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: t X0 = x and let Y0 have the stationary distribution ⇡ , then Yn has distribution ⇡ , and Using (1.14) it follows that X |pn (x, y ) ⇡ (y )| 2P (T > n) ! 0 y proving the convergence theorem. 41 1.7. PROOFS OF THE MAIN THEOREMS* Our next topic is the existence of stationary measures Theorem 1.20. Suppose p is irreducible and recurrent. Let x 2 S and let Tx = inf {n 1 : Xn = x}. µx (y ) = 1 X Px (Xn = y, Tx > n) n=0 deﬁnes a stationary measure with 0 < µx (y ) < 1 for all y . Why is this true? This is called the “cycle trick.” µx (y ) is the expected number of visits to y in {0, . . . , Tx 1}. Multiplying by p moves us forward one unit in time so µx p(y ) is the expected number of visits to y in {1, . . . , Tx }. Since X (Tx ) = X0 = x it follows that µx = µx p. • @ • A • y A A A• @• A A • A A• @ @ Tx x• 0 Figure 1.2: Picture of the cycle trick. Proof. To formalize this intuition, let pn (x, y ) = Px (Xn = y, Tx > n) and ¯ interchange sums to get X µx (y )p(y,...
View Full Document

## This document was uploaded on 03/06/2014 for the course MATH 4740 at Cornell.

Ask a homework question - tutors are online