Authors: Joe F. Hair Jr.; Lucy M. Matthews; Ryan L. Matthews; Marko Sarstedt
Addresses: Marketing and Quantitative Methods, Mitchell College of Business, University of South Alabama, 5811 USA Drive South, Mobile, AL 36688, USA ' Department of Marketing, Jones College of Business, Middle Tennessee State University, MTSU Box 40, Murfreesboro, TN 37132, USA ' RLM Enterprises, L.L.C., 3119 Turret Way, Murfreesboro, TN 37129, USA ' Otto-von-Guericke-University Magdeburg, Chair of Marketing, Universitätsplatz 2, 39106 Magdeburg, Germany; Faculty of Business and Law, University of Newcastle, University Drive, Callaghan, NSW 2308, Australia
Abstract: Numerous statistical methods are available for social researchers. Therefore, knowing the appropriate technique can be a challenge. For example, when considering structural equation modelling (SEM), selecting between covariance-based (CB-SEM) and variance-based partial least squares (PLS-SEM) can be challenging. This paper applies the same theoretical measurement and structural models and dataset to conduct a direct comparison. The findings reveal that when using CB-SEM, many indicators are removed to achieve acceptable goodness-of-fit, when compared to PLS-SEM. Also, composite reliability and convergent validity were typically higher using PLS-SEM, but other metrics such as discriminant validity and beta coefficients are comparable. Finally, when comparing variance explained in the dependent variable indicators, PLS-SEM was substantially better than CB-SEM. Updated guidelines assist researchers in determining whether CB-SEM or PLS-SEM is the most appropriate method to use.
Keywords: structural equation modelling; SEM; PLS-SEM; CB-SEM.
International Journal of Multivariate Data Analysis, 2017 Vol.1 No.2, pp.107 - 123
Available online: 20 Oct 2017 *Full-text access for editors Access for subscribers Purchase this article Comment on this article