Stochastic

# Using this with p zn 1 ezn gives the desired result

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: Having found the martingale it is easy now to conclude: Theorem 5.16. Consider a left continuous random walk with positive mean. Let a < x and Va = min{n : Sn = a}. Px (Va < 1) = e↵(x a) Proof. Again if one argues casually e↵x = Ex (exp(↵Va )) = e↵a Px (Va < 1) but we have to prove that there is no contribution from {Va = 1}. To do this note that Theorems 5.13 and 5.11 give e↵x = E0 exp(↵SVa ^n ) = e↵a P0 (Va n) + E0 (exp(↵Sn ); Va > n) exp(↵Sn ) e↵a on Va > n but since P (Va = 1) > 0 this is not enough to make the last term vanish. The strong law of large numbers implies that on Va = 1, Sn /n ! µ > 0, so the second term ! 0 as n ! 1 and it follows that e↵x = e↵a P0 (Va < 1). 171 5.4. APPLICATIONS When the random walk is not left continuous we cannot get exact results on hitting probabilities but we can still get a bound. Example 5.13. Cram´r’s estimate of ruin. Let Sn be the total assets of e an insurance company at the end of year n. D...
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