This computational study investigates the use of response surfaces to model the
output of the real life processes. In real life, the experimenter does not know the exact
nature of the output of the process. So the output has to be approximated using a
model. This work proposes the generation of a new response surface, Compromise Response
Surface (CRS). The experimenter initially decides on all the possible models to be
considered and the number of design points to be considered per model.
Then the design is generated such that they are equally good for all the models considered. This design is called compromise design. The use of compromise design gives a better design than an
augmented design. Then the methodology for the compromise design generation for linear through cubic for more than two variables is presented. The effectiveness of compromise design for more than two variables is tested on a problem with three variables.