DECOMPOSITION APPROACH TO VARIANCE REDUCTION.

For analyzing stochastic models, simulation trades the tractability problems of analytical techniques for the problem of sampling variability. Variance reduction techniques (VRTs) attack this problem by transforming the simulation experiment in a way that makes it more statistically efficient. Unfortunately, VRTs are infrequently used, even though significant reductions are possible in practical problems. The author introduces some basic concepts of variance reduction, and uses a new taxonomy of VRTs as the basis for an algorithm to select appropriate VRTs for general simulation experiments.