Stata中xtmixed与xtgee相对于纵向数据分析的比较

Differences between Stata Commands Xtmixed and Xtgee on Longitudinal Data Analysis

目的比较Stata统计软件中xtmixed命令与xtgee命令的差别。方法使用Stata9.0统计软件,以纵向观察的青少年牙齿发育的数据为例,从对完全数据、随机缺失的非完全数据和非随机缺失的非完全数据以及命令的应用条件等方面对xtgee命令和xtmixed命令进行了深入比较。结果xtgee命令和xtmixed命令在处理应变量为连续性变量的二水平结构的数据的结果相似,并且在应变量为连续性变量的非完全数据中,不论缺失数据的类型如何,两个命令的结果非常相近。xtmixed可以用于分析多水平结构的数据,而xtgee命令仅能分析两水平结构的数据。结论在分析应变量为连续性变量的二水平结构的数据(完全数据或非完全数据)时,xtgee命令和xtmixed命令结果一致。

Objective To compare Stata command xtmixed with xtgee for longitudinal data analysis. Methods With longitudinal dataset of a dental studY as a example, to compare the two commands in the following three aspects： the results in full dataset, the results in the datasets with missing values （dataset with random missing data and dataset with selective missing data） and applications of the two commands using Stata 9.0 software. Results For the continuous outcome variable, the regression coefficients and standard errors calculated with xtgee and xtmixed command were only slightly different, on both the full dataset and the incomplete datasets. The applications of the two commands are different in some aspects .Conclusion For continuous outcome variable, xtgee and xtmixed are highly comparable on data of one nested level, and both commands lead to ＇ valid＇ results even when applied to incomplete datasets.