Abstract
The response surface technique ridge analysis was developed within the context of unconstrained, or nonmixture, explanatory variables. It has, apparently, been generally considered nonapplicable to mixture data. In this discussion, two methods are introduced for analyzing quadratic mixture response surfaces with ridge analysis. The first may be useful with standard response surface designs (e.g., Draper and Lawrence 1965a,b; Thompson and Myers 1968) and the second offers an alternative method of examining the effects of individual components in constrained designs. Both methods are demonstrated on real data.