Robust designs

Abstract

A response surface design for fitting an empirical interpolation function Y = Xβ + ε can be judged by various criteria, fourteen of which we enumerate. One of these is that the design should be insensitive to wild observations. A measure of such sensitivity is the sum of squares of diagonal elements of the matrix X ( X′X ) −1 X′ , when the number of runs and the number of parameters are both fixed.