探讨具有异方差结构的一阶线性回归模型的DS最优设计问题,给出其异方差结构函数满足一定条件的固定系数一阶线性回归模型的DS最优设计的解析式,并证明随机系数一阶线性回归模型不存在任何DS最优设计.
The paper investigates the DS-optimal designs for first-degree regression models with heteroscedastic structures.Sufficient conditions for the heteroscedastic structure are given to make sure that the equireplicated design is DS-optimal for the first-degree regression models with heteroscedastic errors.It is shown that there exists no DS-optimal design for first-degree random coefficient regression models.