Fractional Factorial Designs for Estimating Main Effects

Abstract

This chapter focuses on efficient designs intended for estimating main effects, including regular resolution III 2^k−f fractional factorial designs, Plackett– Burman and other designs based on Hadamard matrices, nonorthogonal saturated main effect designs, and supersaturated designs. These designs are useful for identifying important factors when it is reasonable to expect that their effects are essentially additive. Even when the assumption of additive effects is suspect, these designs can produce useful initial experiments, provided they are augmented with additional runs. Whenever possible, we will explore evidence for two-factor interactions, even with these screening experiments.