Factor screening methods in computer simulation experiments

The use of a computer simulation model may be viewed as an experiment in which a set of k controllable factors are varied according to an experimental design, and the effect of these factors on the output observed. Usually not all k factors will be active, and considerable efficiency in the use of the model will result if the subset of active factors can be identified. Factor screening methods are useful in identifying the set of active factors. This paper discusses experimental design methods useful for factor screening in computer simulation. The general strategy recommended is group screening methods combined with 2k-p fractional factorial designs. Some variance reduction considerations are also discussed.