【摘要】 <正> 假定预报量y与预报因子x₁,x₂,…,x_n之间有如下的线性关系: y=β₁x₁+β₂x₂+…+β_nx_n+e, （1）其中y为随机变数,x₁,…,x_n为非随机变数,β₁,…,β_n为常系数,e为随机误差,其分布不依赖于x₁,…,x_n,均值为0,方差为σ²。 假定我们已有N组独立的数据,把它们写成矩阵形式,即更多还原

【Abstract】 In this paper, we first point out that the weighted least square method is superiorto the unweighted one when we take the MSE for single prediction as a criterion ofregressor selection. Then, an estimator of the MSE for the weighted least square methodand the corresponding screening scheme are given. Finally, using Hald’s cement data, wecontrast the scheme with the ordinary stepwise regression.更多还原