The use of variance reduction techniques in the estimation of simulation metamodels

Variance reduction techniques can be useful strategies for improving the estimates of simulation metamodel coefficients. Depending upon the goals of the experimenter, the type of metamodel being estimated, and the characteristics of the system being simulated, an appropriate variance reduction technique can be applied. This paper provides a review of recent research that investigates the application of variance reduction techniques in the simulation metamodeling context. One strategy, Schruben and Margolin´s (1978) assignment rule, which utilizes a combination of antithetic and common random number streams, is found to be a particularly useful variance reduction technique for the estimation of simulation metamodels