A Structural Equation Modeling Approach Combining Multitrait-Multimethod Designs with Moderated Mediation Analysis

Researchers who study clinical and developmental psychology are often interested in answering questions such as how interventions work, when treatment begins to improve health outcomes, or for whom treatment has the greatest impact. Answers to these and similar questions impact the general understanding of health and behavior, and can be imperative for effectively implementing intervention and prevention programs. To evaluate such complex relationships among variables, researchers have turned to moderated mediation analysis. Moderated mediation analysis is a statistical tool used to identify the conditional processes among observed or latent variables. However, in developmental and clinical psychology, variables are regularly measured using multiple sources or multiple methods. In fact, best practice recommendations in clinical psychology suggest measuring variables with multiple methods (Achenbach, 2006). The question arises how to use multimethod assessments in statistical analyses such as moderated mediation analysis. The objectives of the present study were to create a multimethod moderated mediation model, apply the model to an extant dataset of child developmental behaviors, and evaluate conditions under which the model performed well using a Monte Carlo simulation study. Results from the application showed that the indirect path from hyperactivity to academic impairment through oppositional defiant behavior was significant but not moderated by inattention. Results from the simulation study indicated that excluding true method effects from a moderated mediation model resulted in unacceptable parameter and standard error bias. These results point to the advantages of using the M4 model to evaluate moderated mediation in the presence of multimethod data