DeJong Trey L, Chen Qi
Department of Mathematics and Statistics, Center for Interdisciplinary Statistical Education and Research, Washington State University, Pullman, WA, United States.
Department of Educational Psychology, The College of Education, University of North Texas, Denton, TX, United States.
Front Psychol. 2023 Jun 27;14:1156962. doi: 10.3389/fpsyg.2023.1156962. eCollection 2023.
To determine which interventions work best for which students, precision education researchers can examine aptitude-treatment interactions (ATI) or skill-by-treatment interactions (STI) using longitudinal multilevel modeling. Probing techniques like the slopes difference test fit an ATI or STI framework, but power for using slopes difference tests in longitudinal multilevel modeling is unknown. The current study used simulation to determine which design factors influence the power of slopes difference tests. Design factors included effect size, number of waves, number of clusters, participants per cluster, proportion of assignment to the treatment group, and intraclass correlation. Of these factors, effect size, number of waves, number of clusters, and participants per cluster were the strongest determinants of power, model convergence, and rates of singularity. Slopes difference tests had greater power in longitudinal multilevel modeling than where it is originally utilized: multiple regression.
为了确定哪些干预措施对哪些学生最有效，精准教育研究人员可以使用纵向多层次模型来检验能力-治疗交互作用（ATI）或技能-治疗交互作用（STI）。像斜率差异检验这样的探测技术符合ATI或STI框架，但在纵向多层次模型中使用斜率差异检验的功效尚不清楚。当前的研究使用模拟来确定哪些设计因素会影响斜率差异检验的功效。设计因素包括效应大小、波数、聚类数、每个聚类中的参与者数量、分配到治疗组的比例以及组内相关性。在这些因素中，效应大小、波数、聚类数和每个聚类中的参与者数量是功效、模型收敛和奇异性发生率的最强决定因素。斜率差异检验在纵向多层次模型中的功效比其最初使用的多元回归更高。
