Institution: Georgia Tech
Street: School of ISyE, Georgia Tech
City: Atlanta, GA
Authors: David Goldsman, School of ISyE, Georgia Tech
Title: Random Walk Methods for Analyzing the Performance of Bernoulli and Multinomial Selection Procedures
Abstract: In this talk, we use random walk and simulation techniques to analyze the performance of certain selection procedures. In particular, we formulate the assessment of these procedures as multi-dimensional gambler`s ruin problems. This formulation allows us to obtain exact (and sometimes closed-form) procedure performance results instead of having to resort exclusively to Monte Carlo simulation methods. We illustrate our work on a number of (a) Bernoulli selection procedures (where we are interested in finding that one of k Bernoulli populations having the largest success probability), and (b) multinomial selection procedures (where we are interested in finding the most probable multinomial cell).