The Performance of Two-Stage Group Screening in Factor Screening Experiments

Abstract

Many experimental situations require consideration of a large number of factors. Often it is anticipated that only a small subset of the factors will be important in explaining the response variable. To identify those factors efficiently, one may conduct a screening experiment. One possible screening strategy is two-stage group screening. A basic assumption of this strategy is that the directions of possible effects are known or are correctly assumed, a priori. In practice, however, this assumption is often unreasonable. This article examines, in the case of zero error variance, the performance of two-stage group screening when this assumption is false. The results provide guidance in using and selecting a group screening strategy.