For all response surface types, Goodness of Fit is calculated for learning points, which are the DOE points and refinement points used to generate the response surface, and for verification points. You can display or hide the verification points in the Table and Chart views by changing the settings in the Properties view. If a response surface has discrete input parameters, you can view the response surface for each combination of discrete levels by selecting the discrete input values in the Properties view.
For Genetic Aggregation response surfaces, Goodness of Fit is also calculated for learning points by using the cross-validation technique. The advantages of cross-validation are that it assesses:
Stability of a response surface without having to calculate additional points (verification points).
Richness of the DOE (instead of only measuring the quality of the fit). When cross-validation produces bad metrics, it means that there are not enough points.
For more information on how Goodness of Fit is calculated for the different types of response surface points, see Goodness of Fit Calculations.