Exploiting correlation in the construction of D-optimal response surface designs

TL;DRAbstract

Cost considerations and difficulties in performing completely randomized experiments often dictate the necessity to run response surface experiments in a bi-randomization format. The resulting compound symmetric error structure not only affects estimation and inference procedures but it also has severe consequences for the optimality of the designs used. Fir this reason, it should be taken into account explicitly when constructing the design. In this paper, an exchange algorithm for constructing D-optimal bi-randomization designs is developed and the resulting designs are analyzed. Finally, the concept of bi-randomization experiments is refined, yielding very efficient designs, which, in many cases, outperform D-optimal completely randomized experiments.

Chat with Paper

AI Agents for this Paper