A cluster analysis selection strategy for supersaturated designs

Supersaturated designs (SSDs) are widely researched because they can greatly reduce the number of experiments. However, analyzing the data from SSDs is not easy as their run size is not large enough to estimate all the main effects. This paper introduces contrastorthogonality cluster and anticontrast-orthogonality cluster to reflect the inner structure of SSDs which are helpful for experimenters to arrange factors to the columns of SSDs. A new strategy for screening active factors is proposed and named as contrast-orthogonality cluster analysis (COCA)method. Simulation studies demonstrate that thismethod performs well compared to most of the existing methods. Furthermore, the COCAmethod has lower type II errors and it is easy to be understood and implemented. © 2010 Elsevier B.V. All rights reserved.