Linda Wänström: Optimal Allocation for Comparing Treatment Effects with Cost Restrictions

This paper considers optimal allocation of respondents to experimental groups when focus is on minimizing a function of the group variances to a fixed total experimental cost. Explicit expressions for A_A-optimal designs, and general expressions for D_A-optimal designs are derived. In general, more respondents should be allocated to groups with larger variances and lower experimental unit costs. If regions of plausible values for the group variances are specified, minimax designs can be constructed by substituting the maximum group variances within the regions into the expressions for the A_A- and D_A-optimal designs, respectively.