Selecting a Best Two-Level 16-Run Screening Design from the Catalog of Nonisomorphic Regular and Nonregular Designs for Six to Eight Factors

When exploring first-order models including two-factor interactions for six to eight factors using a 16-run design, there are many possible model choices. Building on the Johnson and Jones (2010) catalog of the nonisomorphic regular and nonregular design alternatives, we summarize which of these design options are most promising based on two common design criteria. The Pareto fronts based on the criteria E(s(2)) and tr(AA\u27) suggest that only a handful of the possible designs should be considered further, and the best design depends on the relative emphasis given each of the two criteria. This article considers each case of six, seven, and eight factors for 16-run two-level designs and provides numerical and graphical comparisons between the alternatives to highlight the merits of the leading candidates