机构地区: 五邑大学
出 处: 《数理统计与应用概率》 1995年第3期28-34,共7页
摘 要: 本文考虑用附加信息来改进回归方程y_1=X_1β_1+ε_1的系数β_1的LS估计问题,把这一问题归结为一种相依回归(SUR)模型进行讨论,给出了β_1的依赖于样本信息和彼此独立的附加信息的最佳线性无偏估计量,得到了β_1的基于限定残差和非限定残差的两种常用两步估计量的有限样本性质。 For the seemingly unrelated regression equations(SUR):yi = Xiβi+ εi (i = 1, 2, … ,m),the coefficient vector of β1 is considered. On the condition that y2,…,ym are independent with each other, the best linear unbiased estimate of β1 is given as . The finite sample results of two-step estimators of β1based on restricted and unrestricted estimates of ∑ = (σij) arc obtained in this paper.