The relationship between optimal designs for microarray and paired comparison experiments

Current models for two-color microarray experiments are structurally similar to linear models for paired comparisons so that essentially the same designs are optimal in both settings. It is shown how methods for generating efficient multi-factor paired comparison designs can be adapted to microarray experiments. We consider main-effects-only models for factors with arbitrary numbers of levels, which appear to be particularly useful for screening experiments, and models including both main effects and first-order interactions when all factors are binary. Design constructions are presented in a step-by-step manner and are illustrated with examples