Assessing Multilevel Mediation With the Partial Posterior p Value

• In Event: 77.056 - Fairmont Royal York Roundtable Session Seven
  In Roundtable Session: 77.056-3 - Innovative Statistical Modeling and Methods

Abstract

Mediated effects can capture the complex systems operating in educational settings. A variety of tests are available to determine the significance of these effects but there are persistent performance issues in multilevel settings. In this study we developed formulas for and investigated the novel resampling based partial posterior p value (p3 method) to test two-level multilevel mediated effects (i.e.,1-1-1, 2-1-1 and 2-2-1). This test formulates inferences based on composite null hypotheses better tracking the asymmetric sampling distributions that arise when considering the product of two random variables. Similar to prior literature with single level studies, we found that the p3 method consistently outperformed several popular and highly recommended mediation tests in terms of Type I error rate and power.

Authors

• Kyle T. Cox, University of Cincinnati

• Benjamin Kelcey, University of Cincinnati