SIMULTANEOUS FACTOR SCREENING AND REGION REDUCTION IN COMPUTER-SIMULATION EXPERIMENTS

This paper outlines a technique for factor screening and experimental region reduction in the early phases of a computer simulation experimental effort. Given a simulation model with n input variables x(i), i = 1,..., n and m responses y(j), j = 1,..., m, we seek to simultaneously reduce the number of factors to a smaller set x(i), i = 1,..., k where k < n, and to reduce the experimental region for each of the k remaining factors to one-half of its initial range. The combination of these actions reduces the experimental region to as small as 1/2n of its original size and eliminates (n-k) factors from further experimentation.