Preserving projection properties when regular two-level designs are blocked

Abstract

Regular two-level designs are useful and popular screening designs, but if they need to be run in blocks, their projection properties can dramatically deteriorate. Interactions may be fully confounded with block defining contrast(s), causing uncertainty in the identification of active factors. In this paper, we demonstrate alternative ways of blocking two-level regular designs such that their projective properties can be preserved or just weakly affected at the expense of just a small decrease in efficiency. Thereby, we can estimate effects we are normally interested in even if the design is blocked. Common regular two-level designs with 16, 32 and 64 runs are considered.