Optimal Selection of Augmented Pairs Designs for Response Surface Modeling

With reference to second-order polynomial regression models, we consider augmented pairs (AP) designs recently introduced by Morris and address two open problems that he posed. These concern the choice of the initial design and the consequence of rescaling the augmented runs. Although some of our results provide theoretical support to the choice of AP designs as in Morris, some others are quite counterintuitive. We also briefly discuss the effect of blocking on our results.