Minimization of model inadequacy when observations subject to shift in means and variances

Abstract

In this article optimality of experimental design for fitting a lower-order polynomial to a higher order response function for the situation in which observations may be subject to shift in means as well as in variances is considered. It is found that Karson, Manson and Hader‘s (1969) optimum designs provide pro-tection, in some sense, against model inadequacies even when observations are subject to shift in means and variances.

Keywords:

• Response surface
• design variables
• design matrix
• Optimum design