Abstract

Many experimental designs in educational and behavioral research involve at least one level of clustering. Clustering affects the precision of estimators and its impact on statistics in cross-sectional studies is well known. Clustering also occurs in longitudinal designs where students that are initially grouped may be regrouped in the following year and regrouped again in subsequent years. The purpose of this article is to explore the clustering effects of multiple waves of grouping on the effect sizes of downstream outcomes; that is, outcomes that are measured after one or more waves of regrouping. We center our framework on the longitudinal cluster randomized trial in which treatments are initially assigned by clusters (such as classrooms) and students are then regrouped into different clusters in subsequent years. This article illustrates how information on the intraclass correlations in longitudinal studies can be used to correct for biases in the effect sizes and variances, and how it can also be used to adjust the significance test for the effects of longitudinal clustering. This article also provides the first empirical evidence of the magnitude of longitudinal clustering effects using data from a state longitudinal data system.

摘要

教育和行为研究中的许多实验设计都涉及至少一个层次的聚类。聚类影响估计量的精度，并且它对横断面研究中的统计数据的影响是众所周知的。聚类也发生在纵向设计中，最初分组的学生可能在下一年重新组合，并在随后几年再次重新组合。本文的目的是探讨多波分组对下游结果的影响大小的聚类效应；也就是说，在一波或多波重组之后衡量的结果。我们将我们的框架集中在纵向集群随机试验上，其中治疗最初由集群（例如教室）分配，然后学生在随后几年被重新分组到不同的集群中。本文说明了如何使用纵向研究中的组内相关信息来纠正效应大小和方差的偏差，以及如何使用它来调整纵向聚类效应的显着性检验。本文还使用来自州纵向数据系统的数据提供了纵向聚类效应幅度的第一个经验证据。