Experimental Power for Three-Level School-Randomized Studies Probing Mediation

• In Event: AERA Poster Session 7
  In Poster Session: Division D Section 2 Poster Session 2

Abstract

We develop expressions to estimate power to detect mediation effects in three-level school-randomized studies. Mapping the sensitivity of designs is of critical importance because such design sensitivity directly governs the types of evidence researchers can bring to bear on theories of action under common sample sizes. The results provide a set of formulas and software tools that guide researchers in the planning multilevel studies incorporating mediation. We extend tests of mediation that can be used both in the planning and analysis phases including the Sobel test, the joint test, the Monte Carlo interval test, and the partial posterior predictive distribution test. We hope to streamline multilevel mediation power analyses in ways that help researchers understand how to plan studies.

Authors

• Yanli Xie, University of Cincinnati

• Benjamin Kelcey, University of Cincinnati

• Fangxing Bai, University of Cincinnati