We consider the effect of common random numbers (CRNs) among simulated systems in two different multinomial selection procedures; the classical procedure BEM (Bechhofer, Elmaghraby, and Morse) and procedure AVC (all vector comparisons). We examine a simple two population scenario and demonstrate analytically how CRNs affect the probability of correct selection (PCS). In addition, various levels of positive correlation are induced to approximate the effect of using CRNs with complex simulated systems. We present analytical and simulation results to show the effect correlation has on the probability of correct selection (PCS) for each solution procedure. This paper responds to Jim Wilson’s question raised at WSC’96 about how the use of CRNs affects multinomial selection procedures.
John O. Miller, Kenneth W. Bauer Jr.