Power and Optimal Sample Allocation in Cluster-Randomized Trials of Mediation

• In Event: 42.035 - Statistical Power and Missing Data

Abstract

Cluster-randomized trials (CRTs) designed to track upper-level mediation (i.e., 2-2-1) provide an excellent means to investigate the mechanisms through which educational interventions affect student outcomes because these interventions are often implemented at the classroom or school level. Unfortunately, literature regarding design strategies for these studies has been sparse. To address this gap, we derived the power and optimal sample allocation formulas for CRTs designed to track upper-level mediation using the Sobel test and other resampling based tests. Probing these formulas suggests that carefully planned CRTs may be well-positioned to detect mediation effects with typical sample sizes. Study findings expand the ability of educational researchers to plan appropriately powered and efficient studies of upper-level mediation.

Authors

• Kyle T. Cox, University of Cincinnati

• Benjamin Kelcey, University of Cincinnati

• Zuchao Shen, University of Georgia