Beyond Definitive Screening Designs

Gaining competitive advantage requires today’s businesses and industries to innovate and improve at an increasing speed. Modern products and production processes are more complex than ever, so optimizing product formulations/compositions and production process settings requires solving high-dimensional decision problems. An important management science tool to cope with this task is statistical design of experiments. A design is a plan to systematically explore alternative product formulations/compositions and process settings, identify the factors that drive product and process performance/quality, and determine the optimal settings for these factors. Screening designs aim at screening out the most important factors from a long list of potentially important ones. In recent years, a new experimental plan, the definitive screening design, has gained popularity in this context. In this project, I will extend the class of definitive screening designs by constructing complete catalogs of experimental plans with similar characteristics and identifying the best plans within these catalogs. I will also use these plans as building blocks for constructing a new class of experimental plan that fills the gap between ordinary factor screening plans and larger experimental plans for product and process optimization. Success in this project will advance the state of the art in design of experiments and fast-track innovation and quality improvement through better experiments.