Further considerations in A-optimal second-order response surface designs

Abstract

Second-order design of experiments over hypercubic regions is under consideration for the set -up in which the experimenters’ main interest is in estimation of parameters corresponding to higher-order terms of the model. The designs optimal under A-optimality criterion for the situation are obtained. Performances of the derived designs and the designs A-optimal for full model relative to each other are investigated. The derived A-optimal designs are found to have very high efficiency and, for all practical purposes, may be equally good for the full model. Under the set-up the best designs are also derived within the class of symmetric product designs and the class of central composite designs. AMS Subject Classification62K99, 62K05.