Optimal paired comparison designs for factorial and quadratic models (Q1084809)

This paper deals with the problem of constructing D-optimal designs in paired comparison experiments. The techniques employed in paired comparisons to estimate the parameters of interest yield a variance covariance matrix which depends upon the unknown parameters. In order to construct designs, the assumption of no treatment differences is made. This assumption is assumed to hold throughout this paper. A method of construction of D-optimal paired comparison designs for response surfaces is given. Also included are D-optimal designs for a (i) quadratic response surface, (ii) quadratic model with interactions, and (iii) factorial model with main effects and first-order interactions.