Abstract:
|
In this work, we discuss the optimal design problems for comparison of response surfaces, which may be used for discriminating the expected responses between two or multiple experimental groups. In some cases in order to be able to discriminate the response surface models, estimation of nonlinear functions of the unknown parameters in the models may be needed. Then prior information about the unknown parameters in the response surface models would be needed for efficient design of experiments on group discrimination. In this talk, the locally-optimal and Bayesian-optimal designs for comparison of response surface models from two experimental groups on the two-dimensional space are presented, under certain prior information of the unknown parameters. Later cases with multiple experimental groups will also be discussed. The results are illustrated with some examples and the Bayesian optimal designs are compared with the locally optimal designs.
|