Powerful Supersaturated Designs When Effect Directions Are Known

• Activity Number: 507
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Maria Weese*+ and Byran Smucker and David Edwards
• Companies: and Miami University and Virginia Commonwealth University
• Keywords: constrained Var(s) ; Dantzig selector ; forward selection ; Supersaturated design

Abstract

As a criterion for supersaturated designs, we suggest minimizing the variance of all squared pairwise inner products subject to a constraint on the average of these off-diagonal elements of the information matrix. Such designs, which we call constrained Var(s)-optimal, were first mentioned in Booth and Cox (1962) and recently discussed in Weese et al. (2013). We demonstrate that when the sign of each main effect is correctly specialized in advance that these designs tend to have significantly higher power to detect active factors when compared to the standard supersaturated criteria that produce designs with small average off-diagonal elements of the information matrix.