William Li, Principal Investigator
Christopher J. Nachtsheim, Co-Principal Investigator

Model-Robust Designs, Non-isomorphic Orthogonal Designs, and Optimal Blocking Designs

The first object of this research is to construct model-robust designs that include supersaturated designs and response-surface design. Model-robust designs usually consider many different possible models that can increase exponentially with the run size. Thus, the problem is extremely computationally intensive. The second objective is to construct complete catalogs of nonisomorphic designs. Orthogonal designs are the most commonly used experimental designs in practice. The choice of optimal designs depends on two important things—criterion and complete catalogs of the candidate orthogonal designs.

These researchers are using a newly developed efficient algorithm and a theory based on the indicator function to construct complete sets of orthogonal designs with economic run sizes. The third objective, which is a follow-up to the second one, is to find optimal blocking schemes for commonly used orthogonal designs. Blocking is a commonly used technique to reduce unwanted variations in statistics. These researchers’ recent results in complete catalogs of orthogonal designs can greatly facilitate the research in the optimal blocking designs.