Design of Experiments—Complex Designs

Abstract

In this chapter we introduce some additional topics in experimental design beyond those discussed in Chapters 6, 12, and 13. The principle of confounding is used to design efficient experiments having many factors but using only a small subset of all possible treatment combinations. Split plot designs involve placing a restriction on the randomization of treatments to experimental units in order to achieve more precision for comparisons involving levels of one factor in exchange for reduced precision for comparisons involving levels of another factor. We illustrate crossover designs that allow for the estimation of treatment effects that can linger across time periods. We show how to test for interaction in two-way designs having exactly one observation at each treatment combination.