Practical Data Analysis with JMP®

Response Surface Designs

Earlier in the chapter we introduced the idea of a response surface that might be nonlinear. In our examples up to this point, we've tested high and low values of factors with the implicit assumption that the response variable would change in linear fashion. For example, if the response increases when a factor level shifts from low to high we might infer that at an intermediate factor level the response variable would also take on an intermediate value. But what if the response actually curves, so that between the two tested factor levels it actually declines?

It is possible to design an experiment explicitly for the purpose of exploring curvature in the response surface. To illustrate the concept and basic structure ...