Using Weak and Strong Heredity to Generate Weighted Design Optimality Criteria for Response Surface Designs

• Activity Number: 633
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Quality and Productivity
• Author(s): Philip Turk*+ and John Borkowski and Boonorm Chomtee
• Companies: West Virginia University and Montana State University and Kasetsart University
• Keywords: design optimality ; Monte Carlo methods ; response surface methodology ; strong heredity ; weak heredity

Abstract

Weighted D, G, A-efficiencies and the weighted scaled average prediction variance IV- criterion are developed using prior probability assignments to model effects based on the concepts of weak and strong heredity (Chipman 1996). These are four measures of design robustness to prior model misspecification. Numerical examples are presented which involve exact calculation of weighted optimality criteria for the central composite designs and include comparisons for small response surface designs in the hypercube. Monte Carlo methods were then used to estimate weighted optimality criteria, thereby avoiding specification of all weak and strong heredity models. An application of weighted optimality to other sets of reduced models is also discussed.