【作者】 赵晋芳；

【导师】 刘桂芬；

【作者基本信息】 山西医科大学 ， 卫生统计学， 2002， 硕士

【摘要】 在医学研究的各个领域常会遇到重复测量资料的数据分析与设计的问题。由于重复测量数据存在自相关性及随机误差分布于不同层次，实际应用中不宜使用常规分析方法。 本文阐述了重复测量资料的特点，对一般线性模型及线性混合效应模型进行了简要对比；并探讨了重复测量数值型变量线性混合效应模型拟合方法，通过对固定效应、随机效应及协方差矩阵的估计，使重复测量数据得以更合理的分析。线性混合效应模型最早由Laird和Ware(1982)提出，它在分析重复测量资料时非常灵活；同时按照替代法，在一定条件下，线性混合效应模型与两阶段模型可相互转化；此外，线性混合效应模型在对不同方差—协方差结构建模时更加便捷。本文不仅讨论了协方差结构的选择，而且还探讨了如何通过求解混合模型方程而得到参数的估计量。 本文围绕医学实例，分别进行了析因分析、裂区方差分析和线性混合效应模型拟合的对比分析，并对线性混合效应模型拟合不同的方差—协方差结构进行分析。通过对几种不同分析结果的解释，进一步阐明了线性混合效应模型的适用条件及原理。 文中所收集的医学资料，均在计算机上利用SAS软件的MIXED过程进行编程分析，取得了较为满意的分析结果与解释，为线性混合效应模型在医学研究中的应用提供了理论依据。更多还原

【Abstract】 In many fields of medical studies, the analysis and design of repeated measurements data are often come across. Because of correlation among repeated measurements data and its hierarchical error, it is not suitable to use ordinal method.In this paper, the linear mixed-effects model of repeated measurements is discussed, and the repeated measurements data obtain reasonable results by the fixed and random effects along with efficient estimate of covariance matrix. The linear mixed-effects model presented by Laird and Ware(1982) is a more flexible model for repeated measures. Upon substitution, the linear mixed-effects model can be always expressed in terms of the two stage model and vice-versa. In addition, the linear mixed-effects model also provides greater flexibility with respect to the assumed variance-covariance structure.In this paper, alternative covariance structures are discussed,as well as how to obtain estimates of parameters by solving the mixed model equations.All the medical data collected in the paper were analyzed on computer by the PROC MIXED of SAS software system. Satisfactory conclusions and presentations have been obtained. It is wished that the theoretical basis might have beenfulfilled for the popularized application of the linear mixed-effects model in medical studies.更多还原