摘要
目的探讨医学研究中纵向数据广义估计方程(GEE)的准似然独立准则(QIC)分析技术,应用此技术结合医学实例说明。方法采用Stata10.0软件,对GEE进行QIC分析并找出最佳模型。结果比较不同的工作相关结构后,可交换相关结构的QIC值最小,因此被选为最佳结构模型。在此基础上,最佳GEE模型也被选出。结论利用Stata软件,使用该方法通过临床实例的计算,得到的最佳模型更科学和有说服力,其优势在于用QIC标准选择了最佳相关结构和最佳模型。当有更多的协变量进行分析时,比较方程的次数较多,需要探索进一步的简化计算程序。
Objective This study was to apply the Quasi-likelihood under the Independence model Criterion（QIC） technique to the Generalized Estimating Equation（GEE） analysis of longitudinal data in medical research. Methods QIC technique was used to select the best working correlation and best model. Stata 10 software was used for this analysis. This method was applied to an example in clinical. Results After comparison of different working correlation structures, the exchangeable correlation structure has the smallest QIC value and thus was selected as the best correlation structure. Based on the exchangeable correlation structure, the model with covariates central, group and age has the smallest QIC value and thus was identified as the best GEE model. Conclusion This is the first study in China of applying the QIC technique to analyse longitudinal data using GEE in Stata software. The QIC technique ean be used for identifying the best working correlation structure and the best GEE model. However, there are many covariates in the model the times of comparison of different models increase dramatically. Further simplification of calculation program is needed.
出处
《中国卫生统计》
CSCD
北大核心
2008年第4期369-372,共4页
Chinese Journal of Health Statistics
基金
2007年得到国家公派出国留学基金委资助
