Stochastic

# Proof we begin by computing e expxi to do this 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: Theorems 5.13 and 5.11 to conclude EM0 = EMT ^n then we let n ! 1 and argue that EMT ^n ! EMT . Example 5.10. Gambler’s ruin. Let X1 , X2 , . . . Xn be independent with P (Xi = 1) = p and P (Xi = 1) = q = 1 p Suppose 1/2 < p < 1 and let h(x) = (q/p)x . Example 5.3 implies that Mn = h(Sn ) is a martingale. Let ⌧ = min{n : Sn 62 (a, b)}. It is easy to see that ⌧ is a stopping time. Lemma 1.3 implies that P (⌧ < 1) = 1. Again if we argue casually then (q/p)x = Ex (q/x)S (⌧ ) = (q/p)a P (S⌧ = a) + (q/p)b [1 P (S⌧ = a)] (5.11) Solving gives Px (S⌧ = a) = (q/p)b (q/p)b (q/p)x (q/p)a (5.12) generalizing (1.22). To provide a proof for (5.11), we use Theorems 5.13 and 5.11, to conclude that (q/p)x = Ex M⌧ ^n = (q/p)a P (⌧ n, S⌧ = a) + (q/p)b P (⌧ n, S⌧ = b) + E ((q/p)Sn ; ⌧ > n) 169 5.4. APPLICATIONS Since P (⌧ < 1) = 1 we have P (⌧ n, S⌧ = c) ! P (S⌧ = c) for c = a, b. To handle the third term, we note that since p > 1/2, (q/p)x (q/p)a for a < x < b and hence E ((q/p)Sn ; ⌧...
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