群体相关下的一种回归分析

北京大学学报（自然科学版）

群体相关下的一种回归分析

杨宝慧,孙山泽

北京大学数学科学学院概率统计系，北京，100871

收稿日期:1997-09-03 出版日期:1999-01-20 发布日期:1999-01-20

Regression Estimation Using Cluster Correlation

YANG Baohui, SUN Shanze

Department of Probability and Statistics, School of Mathematical Sciences, Peking University, Beijing, 100871

Received:1997-09-03 Online:1999-01-20 Published:1999-01-20

摘要/Abstract

摘要： 作连续调查研究抽样时经常用到回归估计。如果前后两个样本相互独立，并且它们所包含的元素不同，就无法直接计算出回归系数。本文在群体相关模型下给出了一种新的回归估计。模拟研究表明，当估计总体均值时，如果前后两个样本都来自大的总体，则新的回归估计与样本均值几乎一样好；如果前期样本来自大的总体，而后期样本来自一子总体，则在本文模拟的例子中，新的回归估计比样本均值要好。

关键词: 群体相关, 回归估计

Abstract: Some troubles emerge when we use regression estimator in such a case: in a continuous survey research, when the prior sample and the present sample are independent and they have different elements, we can not compute the regression coefficient. In this paper, we give a new regression estimator under the cluster correlation model. Simulation result shows that, in estimating the population mean, the new estimator is as good as the sample mean under the condition that the two samples are all from the whole population, and better than the sample mean when the prior sample is from the whole population and the present sample is from a subpopulation.

Key words: cluster correlation, regression estimator

中图分类号: