摘要
目的 应用随机系数发展模型与协方差模式模型分析社区卫生服务中心纵向数据,探讨纵向数据分析的问题,为社区随访数据处理提供科学方法.方法 使用R软件对社区卫生服务中心糖尿病重复测量数据分别拟合随机系数发展模型,协方差模式模型以及传统线性回归模型,并比较3种模型的分析结果.结果 随机系数发展模型和协方差模式模型的分析结果与传统线性回归模型不同,2模型较传统线性回归更多的考虑了数据的变异来源.随机系数发展模型与协方差模式模型变量的估计系数结果相近,2者在固定效应的估计上区别往往不是很大,2模型相比,信息标准统计相差也不大.随机系数发展模型倾向于解释组间随机效应,协方差模式模型更关注组内观测之间的联系.R软件nlme package相比于SAS proc mixed,其相应的结果比较与可视化的函数使用更为灵活方便,同时GLS函数提供更多的组内方差协方差模式以供选择.结论 随机系数发展模型与协方差模式模型都能较好的处理重复观测数据组内相关性的问题.2者处理组内相关性的出发点不同,如果强调组内观测之间的联系性,则选择协方差模式模型.相反,如果更关注组间的异质性,强调组间的随机效应,则选择随机系数发展模型.R软件nlme是比较完善的处理混合结构数据的分析包.
Objective To explore the application of random coefficient growth model and covariance pattern model based on longitudinal data of community health service center and provide scientific statistical method for follow up data processing of community. Methods Random coefficient growth model, covariance pattern model and general linear regression model were ap- plied through R and results were compared with each other. Results The results of random coefficient growth model and covari- ance pattern model are similar including information criteria, but different from that of general linear regression model in coeffi- cient estimation of model or fixed effects. They consider the data source of variation more. Random coefficient growth model to covariance pattern model, the former tend to explain random effects between groups, The latter pay more attention to the contact between the observations in a group. It is more flexible and convenient using corresponding function of visual and comparing of nlme package for R than that in SAS proc mixed. Also, the GLS fuction provides more variance-covariance structures among in- tra-individual to choose from. Conclusion Both the random coefficient growth model and covariance pattern model considers the intra-individual correlation in two ways. Random coefficient growth model is apt to account for between-group random effects, and eovariance pattern model place extra emphasis on intra-individual correlation. Nlme package for R is useful to dealing with mixed structure data.
出处
《中国医院统计》
2012年第3期165-168,共4页
Chinese Journal of Hospital Statistics
关键词
纵向数据
随机系数发展模型
协方差模式模型
R软件
Longitudinal data Random coefficient growth model Covariance pattern model R software
