Power and Design in Clustered Regression Discontinuity Designs Probing Mediation

• In Event: 31.038 - Power Analysis in Multilevel Modeling

Abstract

We develop closed-form expressions to estimate the statistical power to detect indirect/mediation effects in clustered regression discontinuity designs. Mapping the sensitivity of such designs is of critical importance because it directly governs the types of evidence researchers can bring to bear on theories of action under different designs and sample sizes. The results provide a set of formulas and software tools intended to inform and guide researchers in planning mediation studies with multilevel regression discontinuity designs. Our analyses consider prospective design-based assessments of indirect effects including the asymptotic Sobel test, component-wise joint test, the resampling-based Monte Carlo interval test, and the resampling-based partial posterior test designed for composite hypotheses. The framework is implemented in the R package BLINDED.

Authors

• Benjamin Kelcey, University of Cincinnati

• Kyle T. Cox, University of Cincinnati