Modelling Double-Moderated-Mediation & Confounder Effects Using Bayesian Statistics

Georgios Chrysochoidis, Lars Tummers, Rens Van de Schoot

Research output: Chapter in Book/Report/Conference proceedingConference contribution


This study provides an example on how to conceptualize and estimate models when double moderated mediation with nominal and continuous (Likert type) variables need to be simultaneously accounted for, and also how to appease reservations given the biases due to the implicit sequential ignorability assumption (endogeneity) regularly overseen in marketing research. We explain the issues and apply the proposed solution using empirical data.
The benefits for research are considerable as this approach is superior to other approaches (e.g. splitting the sample by the binary moderator and estimating a moderated mediation model) while also accounting for accounted confounders.
Original languageEnglish
Title of host publicationAcademy of Marketing 2014
Subtitle of host publicationBest Paper Award in Marketing Research & Research Methodology
Place of PublicationBournemouth
Publication statusPublished - 8 Jul 2014

Cite this