On the relative efficiency of the best second order proportional ANOVA- designs (Q923566)

This paper considers the A-optimal design problem for K-way layouts under the absence of interactions. The so-called approximate designs are interpreted as exact designs. For arbitrary parameter functionals the best second order proportional designs are characterized. These designs turn out to be A-optimal in some interesting cases and to have a high relative efficiency in general. In addition to general optimality results, the paper considers results on (i) designs for main effects and overall mean, (ii) designs for contrasts and a full control treatment mean, and (iii) designs for contrasts and means.