Families of search designs for simultaneous detection of the active main and 2-factor interaction effects

Abstract

MEP and MEP.k have been designated for 2 f factorial experiments with at least f and 2f runs, respectively, to estimate all main effects, while the sparsity principle implies that only a small number of factors are active. A supersaturated design is superior in this regard because it focuses on active factors and saves runs. In this paper, we consider the problem of searching for and estimating k 1 and k 2 non-zero main and 2-factor interaction effects, respectively, that are not known a priori. A family of designs has been constructed and given for 2 f factorial experiments with f runs for f≥11, k 1 =2 and k 2 =1. These designs are able to estimate all possible models consisting any set of at most 4 main and 2 2-factor interaction effects. It is shown that the obtained designs are near D-optimal. © 2019 Elsevier B.V.