CFE 2019: Start Registration
View Submission - CMStatistics
Title: Two-level block designs under model uncertainty Authors:  Pi-Wen Tsai - National Taiwan Normal University (Taiwan) [presenting]
Steven Gilmour - KCL (United Kingdom)
Abstract: Two-level designs are widely used for screening experiments with the goal to identify the few active factors which have major effects. Blocking is a common technique when there is systematic variation among experimental units. Most work on two-level block designs focuses on the method of replacement for finding the best blocking scheme of regular and irregular factorial designs, and we are forced to sacrifice the estimations of one or more factorial effects where all main effects are orthogonal to blocks. We discuss a model-robust block criterion which respects experimenters' prior knowledge on the importance of each effect. It is shown that several minimum aberration criteria for block designs are limiting cases of this model-robust block criterion. Additionally, a coordinate-exchange algorithm is developed to generate new classes of block designs under model uncertainty. We will demonstrate that by relaxing the requirement of orthogonal blocking, more appropriate designs can be recommended under different experimenters priors.