摘要
目的探讨现场调查中多重应答资料的统计推断方法。方法根据多重应答资料不同选项间既存在关联,又不可互换的特点,将其视为多因变量的多元logistic回归问题,采用广义估计方程建模分析,并通过实例及其SAS编程的实现过程进行阐述。结果广义估计方程可以很好地分析自变量对于多重应答资料各项的影响,并能进一步回答各选项的基线应答概率、作业相关矩阵信息;在此模型下,还可以进一步分析各影响因素对于每一选项的具体影响,或各自变量的分层比较具体差异的信息。结论广义估计方程可以作为现场调查中多重应答资料的推断分析。
Objective To establish a statistical methods for analyzing the multiple response data in the field investigation. Methods According to the characteristics that different options in the multiple response data are not only interconnected, but also relative independent, the multiple response data can be considered as a kind of data which could be analyzed by multivariate logistic regression. The data can be analyzed by building the model of generalized estimating equations （GEEs）. The statistical method was illustrated by an example and its SAS programming in detail. Results GEEs can analyze the impact of independent variables on each option of the multiple response data,and can also discover the baseline rate of each option in the multiple response data and the information of work correlation matrix. The specific impact of each factor on each option or the specific differences among stratified comparison of each independent variable could be analyzed by using this model. Conclusion The GEEs can be used as a method to analyze multiple response data in the field investigation.
出处
《中国卫生统计》
CSCD
北大核心
2009年第6期595-598,共4页
Chinese Journal of Health Statistics
关键词
多重应答资料
多因变量
多元LOGISTIC回归
广义估计方程
Multiple response data
Multi-dependent variable
Multivariate logistic regression
Generalized estimating Equations
