Carr G J, Chi E M
Biometrics and Statistical Sciences Department, Procter & Gamble, Cincinnati, OH 45241-2422.
Stat Med. 1992 Jun 15;11(8):1033-40. doi: 10.1002/sim.4780110805.
Various techniques are available for the analysis of repeated measures data, and the appropriate choice depends on distributional assumptions and study design features. A correct analysis must account for potential dependence between repeated observations on the same subject. Liang and Zeger proposed a more unified approach to the analysis of repeated measures data based on the application of generalized estimating equations. We examine the application of these methods to several types of data in which one estimates the mean response directly for each combination of discrete covariates, and uses an identity link. Computations for fitting this type of model are exceptionally simple. Numerical examples suggest that the proposed approach yields estimation and hypothesis testing results consistent with more specialized methods.
有多种技术可用于重复测量数据的分析，而合适的选择取决于分布假设和研究设计特征。正确的分析必须考虑同一受试者重复观测值之间的潜在相关性。梁和泽格基于广义估计方程的应用，提出了一种更统一的重复测量数据分析方法。我们研究了这些方法在几种类型数据中的应用，在这些数据中，人们直接估计离散协变量每种组合的平均响应，并使用恒等链接函数。拟合这类模型的计算非常简单。数值示例表明，所提出的方法产生的估计和假设检验结果与更专门的方法一致。
