通常普通计算器回归只是残差平方和达极小意义下的一个线性无偏估计 ,不是线性无偏最小方差估计。只有当残差平方和中取最优加权时 ,才是线性无偏最小方差估计。本文介绍一种由初始估计 (不加权 )而获得较优权重 (可通过迭代获得最优权重 )的迭代最小二乘法 ,能方便快捷地在普 Usually the ordinary calculator regression is only a linear unbiased estimate in the sense that the sum of residual squares reaches a minimum, and it is not a linear unbiased minimum variance estimate. Only when the residual sum of squared sums is optimally weighted is the linear unbiased minimum variance estimate. This paper presents an iterative least-squares method that obtains better weights from initial estimates (without weighting) that can be optimally weighted by iterations, and can be easily and quickly