Beckett L A, Tancredi D J, Wilson R S
Division of Biostatistics, Department of Epidemiology and Preventive Medicine, University of California, Davis, CA 95616, USA.
Stat Med. 2004 Jan 30;23(2):231-9. doi: 10.1002/sim.1712.
Longitudinal studies offer us an opportunity to develop detailed descriptions of the process of growth and development or of the course of progression of chronic diseases. Most longitudinal analyses focus on characterizing change over time in a single outcome variable and identifying predictors of growth or decline. Both growth and degenerative diseases, however, are complex processes with multiple markers of change, so that it may be important to model more than one outcome measure and to understand their relationship over time. We consider random effects models for the association between the trajectories of two outcomes over time, and then compare their properties to approaches based on separate ordinary least-squares estimates of change. We then illustrate with an example from the Religious Orders Study, a longitudinal cohort study of more than 900 members of Catholic religious orders who have had up to eight annual clinical examinations.
纵向研究为我们提供了一个机会，来详细描述生长发育过程或慢性病的进展过程。大多数纵向分析都集中在刻画单个结果变量随时间的变化，并识别增长或衰退的预测因素。然而，生长和退行性疾病都是具有多个变化指标的复杂过程，因此对多个结果指标进行建模并理解它们随时间的关系可能很重要。我们考虑随机效应模型来研究两个结果随时间的轨迹之间的关联，然后将它们的性质与基于单独的普通最小二乘法变化估计的方法进行比较。然后，我们以宗教团体研究中的一个例子进行说明，该研究是一项纵向队列研究，对900多名天主教宗教团体成员进行了多达八次年度临床检查。
