An Equivalence Testing Approach for Evaluating Substantial Mediation

Abstract

In the past, researchers often used the nonsignificance of the direct path from the predictor to the outcome, in conjunction with a significant indirect effect, to make claims regarding \IeC {\textquoteleft }full mediation\IeC {\textquoteright }. However, the nil hypothesis (i.e., full mediation) is not realistic and it is well known that a nonsignificant test statistic cannot be used to establish the accuracy of a research hypothesis. In this paper, we discuss equivalence testing based procedures for assessing when a mediator explains a substantial proportion of the relationship between a predictor and an outcome. Monte Carlo simulations are used to evaluate the performance of the proposed procedure and compare it against competing alternatives, including traditional tests of full mediation and a proportion mediated approach. The proposed equivalence testing based procedures and the proportion mediated approach performed similarly across the conditions investigated. Recommendations are provided for deciding among the approaches.

Keywords

• mediation
• equivalence testing
• substantial mediation