| Abstract: | Continuing the author's earlier work 6] a method is described which requires that certain experimental runs of a central composite, second-order, response surface design be repeated, thereby providing a more general estimate of the experimental error, at the same time providing more reliable estimates of the effects. The partial duplication of the factorial portion as well as the partial duplication of the star portion has been considered and described. The response surface designs with the star portion duplicated seem to have more potential than the designs with their factorial portions duplicated or partially duplicated. |
