Wang Lijuan Peggy, Maxwell Scott E
Department of Psychology, University of Notre Dame.
Psychol Methods. 2015 Mar;20(1):63-83. doi: 10.1037/met0000030.
This article extends current discussion of how to disaggregate between-person and within-person effects with longitudinal data using multilevel models. Our main focus is on the 2 issues of centering and detrending. Conceptual and analytical work demonstrates the similarities and differences among 3 centering approaches (no centering, grand-mean centering, and person-mean centering) and the relations and differences among various detrending approaches (no detrending, detrending X only, detrending Y only, and detrending both X and Y). Two real data analysis examples in psychology are provided to illustrate the differences in the results of using different centering and detrending methods for the disaggregation of between- and within-person effects. Simulation studies were conducted to further compare the various centering and detrending approaches under a wider span of conditions. Recommendations of how to perform centering, whether detrending is needed or not, and how to perform detrending if needed are made and discussed.
本文拓展了当前关于如何使用多层模型通过纵向数据区分个体间效应和个体内效应的讨论。我们主要关注中心化和去趋势化这两个问题。概念性和分析性工作展示了三种中心化方法（无中心化、总均值中心化和个体均值中心化）之间的异同，以及各种去趋势化方法（无去趋势化、仅对X去趋势化、仅对Y去趋势化以及对X和Y都去趋势化）之间的关系和差异。提供了两个心理学领域的实际数据分析示例，以说明使用不同的中心化和去趋势化方法区分个体间效应和个体内效应时结果的差异。进行了模拟研究，以在更广泛的条件下进一步比较各种中心化和去趋势化方法。文中给出并讨论了关于如何进行中心化、是否需要去趋势化以及如果需要如何进行去趋势化的建议。
