ABSTRACT

Studying development processes, as they unfold over time, involves collecting repeated measures from individuals and modeling the changes over time. One methodological challenge in this type of longitudinal data is separating retest effects, due to the repeated assessments, from developmental processes such as maturation or age. In this article, we describe several specifications of latent change score models using age as the underlying time metric and include parameters to account for retest effects. We illustrate the models with data on fluid reasoning collected from children and adolescents in a cohort-sequential design ranging from 6 to 20 years. Our models include alternative approaches to specify retest effects at the structural or measurement level of the model, and as an observed or a latent covariate. We discuss the benefits and limitations of the different approaches for univariate and multivariate data in the context of studying developmental processes.

摘要

研究发展过程，随着时间的推移而展开，包括从个人那里收集重复的测量值，并对随时间的变化进行建模。此类纵向数据中的一个方法学挑战是将由于重复评估而导致的重新测试效果与成熟或年龄等发育过程分开。在本文中，我们描述了使用年龄作为基础时间度量的潜在变化评分模型的几种规范，并包括用于解释重新测试效果的参数。我们使用从 6 到 20 年的队列顺序设计中从儿童和青少年收集的流体推理数据来说明模型。我们的模型包括在模型的结构或测量级别指定重新测试效果的替代方法，并作为观察或潜在协变量。