Double Moderated Mediation Models: Problems and (Part) Remedies

Chryssochoidis, George (2018) Double Moderated Mediation Models: Problems and (Part) Remedies. Journal of Modelling in Management, 13 (1). pp. 50-80. ISSN 1746-5664. (doi:https://doi.org/10.1108/JM2-06-2016-0053) (Access to this publication is currently restricted. You may be able to access a copy if URLs are provided)

PDF - Author's Accepted Manuscript
Download (810kB) Preview
[img]
Preview
PDF - Publisher pdf
Restricted to Repository staff only
Contact us about this Publication Download (810kB)
[img]
Official URL
http://dx.doi.org/10.1108/JM2-06-2016-0053

Abstract

Purpose: Researchers in management regularly face modelling issues that involve double moderated mediation models. Here, we illustrate how to conceptualise, specify and empirically estimate mediation effects when having to simultaneously account for continuous (Likert type) and nominal (i.e., group) moderator variables. Researchers’ estimates of the mediation effects suffer serious bias due to the effects of unaccounted confounders. This is an issue that plagues management research and this work shows how to address these valid reservations for our focus models. In aiming to inform a wider management audience, we deliberately use the rich context of a focus case since this allows us to clarify the nuances that management researchers face applying double moderated mediation models. Specifically, our focus case is on professionals’ willingness to implement a new government policy. We also combine traditional and Bayesian statistical approaches and explain the differences in estimation and interpretation that are associated with the Bayesian approach. Explaining, and exemplifying the use of, the models we focus on can substantially increase the robustness of the methods employed in management research and can considerably improve the quality of the generated theoretical insights. We also clarify important assumptions and solutions. Design/methodology/approach: We employ a doubled moderated mediation Bayesian approach. We draw our sample data from a population of 5,199 professionals, all members of either the Dutch Association of Psychologists (NIP) or the Dutch Association for Psychiatry (NVvP). The data collection process resulted in 1,307 questionnaires being returned; a response rate of 25%. All the items were measured using a Likert scale, ranging from ‘strongly disagree’ to ‘strongly agree’, unless stated otherwise. Findings: Explaining, and exemplifying the use of, the models we focus on can substantially increase the robustness of the methods employed in management research and can considerably improve the quality of the generated theoretical insights. Research limitations/implications: Practical implications: Originality/value: This is an original approach exemplified for wider use by management researchers.

Item Type: Article
Uncontrolled keywords: Moderated mediation, Bayesian, Sequential ignorability
Divisions: Faculties > Social Sciences > Kent Business School
Faculties > Social Sciences > Kent Business School > Marketing
Depositing User: George Chryssochoidis
Date Deposited: 04 Jan 2017 13:16 UTC
Last Modified: 13 Apr 2018 10:38 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/59775 (The current URI for this page, for reference purposes)
Chryssochoidis, George: https://orcid.org/0000-0001-9868-7119
  • Depositors only (login required):

Downloads

Downloads per month over past year