分层整群随机抽样数据的不同分析方法及结果比较  

Comparison of Different Methods for Analyzing Data Obtained from Stratified Cluster Random Sampling

作　　者：陈丹萍[1] 赵耐青[1] 林燧恒[1] 

机构地区：[1]复旦大学公共卫生学院卫生统计和社会医学教研室,200032

出　　处：《中国卫生统计》2010年第2期122-124,128,共4页Chinese Journal of Health Statistics

基　　金：上海市重点学科建设项目资助(项目编号:B118)

摘　　要：目的探讨分层整群抽样数据应用SAS9.1分析时,不同分析方法对结果的影响。方法比较多因素logistic回归,surveylogistic回归以及广义线性混合效应模型(glimmix)在分层整群抽样数据中的统计分析,并用实例加以说明。结果不同的方法所产生的结果是有差别的。surveylogistic回归与广义线性混合效应模型对模型中各回归系数的标准误进行了调整,使得其比多因素logisitic回归中的标准误大。在实例分析中各危险因素的OR值也发生了变化,其95%可信区间都有不同程度的增宽。结论在分层整群抽样中,为减少模型系数标准误估计的向下偏倚以及第Ⅰ类错误的发生,surveylogistic回归与广义线性混合效应模型都是比较适用的,不建议使用多因素logisitic回归。Objective To explore the results of different statistical methods for analyzing data obtained by stratified cluster random sampling.Methods Multivariable logistic regression,surveylogistic regression and generalized linear mixed models（glimmix） in SAS9.1 were compared for analyzing data obtained by stratified cluster random sampling.Results When surveylogistic and GLIMMIX were used,the standard errors of the regression coefficients were larger than those from multivariable logistic regression.The odds ratios（ORs） of the risk factors also showed small changes,and the 95% confidence intervals of the ORs were wider.Conclusion To minimize the downward bias of the standard errors of the regression coefficients and the occurrence of typeⅠerror,surveylogistic regression and glimmix are more appropriate than multivariable logistic regression for analyzing data obtained by stratified cluster random sampling.

关 键 词：分层整群随机抽样 surveylogistic回归 广义线性混合效应模型 

分 类 号：R195[医药卫生—卫生统计学]