研究了多响应回归模型下含有异常观测值的试验设计问题,导出了衡量设计稳健性的准则.以中心组合设计为例,给出了角点和中心点次数不同时,使得稳健性准则达到最小的星点臂长,从而找到了相应的稳健设计.
We consider the design problem of retrieving the response surface if each response is disturbed by an outlier, in the context of multi -response regression models. We aim at finding a robust design that is insensitive to outliers of observations. The theoretical framework is illustrated by an example with quadratic polynomial regression and under a central composite design.