Sensitivity of Design Optimality Criteria to Second-Order Response Surface Model Restrictions and Center Point Replications

Abstract

In this work, we investigate the role of several model characteristics and center point replications on the properties of A-, D-, G-, andIV- optimality for four –factor second-order response surface design. It was discovered thatA-, D-, and G- efficiencies tend to reduce as the center points are replicated while the effect on scaled average prediction variance tends to be very insignificant. Among the restricted models consid ered, the pure linear model (model1) turns out to be the best in terms of quality of estimation and model prediction.