Abstract
Optimization of a pharmaceutical production process requires intensive screening to determine which factors from a predetermined list affect the properties of the product. The screening is called intensive because two-factor interactions (2fi) are likely to occur. Therefore, identification of main effects in an intensive screening situation requires a statistical design that permits estimation of these effects independent from estimation of 2fi. A recently completed catalog offers many two-level designs with 32, 40, and 48 runs that are suitable for this purpose. However, there may be practical reasons to divide the runs of a design into two equally sized blocks. In this article, I consider optimum blocking for the two-level designs in the catalog. This article has supplementary materials online.
Acknowledgments
This work has been performed under the framework of the Dutch Top Institute Pharma (project D6-203). The author thanks Uwe Thissen and Erik Gout for the discussions that started the problem formulation. The delete-one-factor projections were the result of joint work with Robert W. Mee. Thanks are also due to two referees and an associate editor, whose comments substantially improved the organization of the article.
Notes
†ΔF 3(8) = 18, D = 0.91.