Response Surface - In this type of nonlinear model the experimental response is fitted to a hyper-plane whose coordinates are the main factors (with quadratic and linear interaction terms) remaining in a regression. The usual objective is either to determine if a stationary point in response occurs within the experimental range or to predict the direction in which a stationary point may occur.
Response Surface Analyses should always be used with Contour plots. The algorithm is capable of finding only one stationary point for the data set. Because the stationary point is global for the data, it is possible for a global maxima to be found outside the data range while a local minima exists within the data range.
If there are sufficient data, a response surface may be fitted to any quadratic polynomial. Contour plotting may be used to graphically represent (in 2-D) the response surface.
Factors must have 3 or more levels to be included in a Response Surface Analysis.
Learn more about the Regression tools in Six Sigma Demystified (2011, McGraw-Hill) by Paul Keller, in his online Regression short course (only $99), or his online Black Belt certification training course ($875).