Model Selection for Six-Factor No-Confounding Designs (2019-US-EPO-205)
Aug 28, 2019 5:24 AM
| Last Modified: Oct 21, 2019 9:03 AM
Carly Metcalfe, Research Assistant, Arizona State University Bradley Jones, JMP Distinguished Research Fellow, SAS Douglas Montgomery, Regents' Professor, Arizona State University
Nonregular designs are a preferable alternative to regular resolution IV designs because they avoid confounding two-factor interactions. As a result, nonregular designs can estimate and identify a few active two-factor interactions. However, due to the sometimes complex alias structure of nonregular designs, some classic screening strategies fail to identify all active effects. In this poster we highlight several no-confounding, six-factor, 16-run nonregular designs with orthogonal main effects. We propose an all-possible-models selection approach that selects the minimum mean square error model from all fifteen second-order, four-factor models. We show that by using our proposed method the probability of missing active effects is minimal for a large variety of hypothetical models. A simulation using JSL is performed to compare this method against a forward selection stepwise approach.