Optimal designs for treatment-control comparisons in the presence of two- way heterogeneity (Q1062398)

In this paper we shall consider the additive 3 factor linear model such as is used in Latin square design settings. We seek optimal designs under this model for comparing p test treatments each simultaneously with a control. Conditions under which a design is A-optimal and under which a design is E-optimal are derived and examples given of when these conditions are satisfied. Since E-optimality is a criterion that involves the variances of contrasts other than those of test treatments vs. the control, we show that many E-optimal designs are also optimal under two criteria involving only the variances of test treatments vs. the control contrasts.