Title of the talk: Definitive screening as a system for experimental design
The use of Definitive Screening Designs (DSDs) (Jones and Nachtsheim, 2011) has grown rapidly in recent years. Applications include catalyst development (Olsen, et al., 2014), drug synthesis (Renzi, et al., 2014), biological process optimization (Kauffman, et al., 2015, Erier et al., 2013), carbon synthesis (Libbrecht et al., 2015) and optimization of mass spectrometry (Hecht, et al., 2015). DSDs are small, three-level, orthogonal designs that can be used for efficiently screening main effects, but because active two-factor interactions (2FIs) and quadratic effects can be identified, they have been increasingly used for simultaneous screening and optimization of a system. In this sense they are sometimes referred to as ``one-step RSDs.'' In this presentation, we first provide an overview of DSDs and their properties.. We then describe briefly (1) how categorical factors can be accommodate; (2) how DSDs can be flexibly blocked; and (3) how to augment a DSD for estimation of a full quadratic model and/or for the search for optimum operating conditions. Finally we will introduce a new, effective model selection procedure that leverages the orthogonality and fold-over structure of the designs.
Dr. Nachtsheim is a professor and the Frank A. Donaldson Chair of Operations Management in the Supply Chain and Operations Department in the Carlson School of Management at the University of Minnesota. Dr. Nachtsheim received and his Ph.D. in Operations Research from the University of Minnesota, served as staff member in the Statistics Group at Los Alamos National Laboratory from 1978-1981, and as Senior Statistician at General Mills from 1982-1984. In 1984 he joined the University, serving as Chair in the Department of Operations and Management Science from 1993-1996 and Associate Dean of Faculty and Research from 1996-2000. Dr. Nachtsheim's teaching and research interests center on the optimal design of industrial experiments, regression and predictive modeling, and quality management. Among his major publications are two texts: Applied Linear Statistical Models, 5th Edition, 2005, Richard D. Irwin, and Applied Linear Regression Models, 4th Edition, 2004, Richard D. Irwin (both with John Neter, Michael Kutner, and William Li).
Contact Name: Ron Christensen
Contact Email: firstname.lastname@example.org