The Impact of Ignoring the Treatment-Mediator Interaction on Traditional Tests of Indirect Effects

• In Event: 47.072 - Model Contemplations: Identification, Detection, Effects, and Errors

Abstract

Mediating relationships are commonly hypothesized in the social and behavioral sciences. Using the counterfactual framework, direct and indirect effects can be estimated when interactions and nonlinearities are also included in the mediation model. These estimates are not well-known among applied researchers in the social and behavioral sciences. Nonetheless, interactions between treatment and mediator variables are regularly recognized in causal mediation analysis. A simulation study was conducted to examine the impact of ignoring the treatment-mediator interaction on the performance of traditional tests of indirect effects and the relative parameter bias of estimates. Preliminary results indicated that relative parameter bias associated with parameter estimates, including the indirect effect, was unacceptable when excluding the interaction and parameters were consistently overestimated in these models.

Authors

• Tiffany Ann Whittaker, The University of Texas at Austin

• Pierce Cappelli, The University of Texas at Austin