回归系数的极小极大估计

Minimax Estimation for the Regression Coefficients

将文献[1]中提出的求不相容方程组的极小极大解的方法,应用于线性回归方程中的参数估计,提出了一个包括两个普通的最小二乘法的算法。在基于"最大残差最小"的意义下,得到的参数估计优于最小二乘估计。

In the paper the method of extracting minimax solution from inconsiste nt linear equations raised in reference is applied to the parametric estimation of linear regression, and a algorithm including two common least squares method s is advanced. In the view of minimax residual, obtaining parametric estimation is superior to the estimation by least squares method.