TITLE:
Structural Equation Modeling in Social Sciences: A Comparative Guide to Covariance-Based and Partial Least Squares Approaches
AUTHORS:
Nammuni Arachchige Nuwan Janaka Maduwansha
KEYWORDS:
Structural Equation Modeling, Covariance-Based SEM, Partial Least Squares SEM, Goodness-of-Fit, Validity, Reliability
JOURNAL NAME:
Open Journal of Social Sciences,
Vol.14 No.11,
November
11,
2026
ABSTRACT: Structural equation modeling (SEM) is one of the most widely used multivariate techniques in the social and behavioral sciences, allowing researchers to test complex networks of relationships among observed and latent variables simultaneously, including direct, mediating, and moderating effects. Despite its popularity, researchers frequently struggle to distinguish between the two dominant estimation traditions, covariance-based SEM (CB-SEM) and partial least squares SEM (PLS-SEM), and to apply the correct evaluation criteria for each. This article provides a pedagogically oriented, comparative overview of SEM intended for social science researchers. It defines SEM and its measurement and structural components, and explains how these components extend beyond simple direct-effects models to accommodate mediation analysis, in which an intervening variable transmits or partially transmits the effect of a predictor on an outcome, and moderation analysis, in which the strength or direction of a relationship depends on a third, conditioning variable. The article then contrasts the theoretical assumptions and objectives of CB-SEM and PLS-SEM, including their differing suitability for testing indirect and conditional effects, and consolidates the recommended goodness-of-fit, normality, reliability, and validity criteria, together with their accepted threshold values, for each approach. It further outlines how mediating and moderating effects are estimated and assessed under each approach, sets out separate assessment criteria for reflective and formative measurement models, distinguishes in-sample predictive relevance from out-of-sample prediction, and cautions that directional paths estimated from observational data do not by themselves establish causal effects. By synthesizing guidance from foundational and contemporary methodological sources, the article aims to serve as a practical, citation-supported reference for researchers designing, estimating, and reporting SEM-based research, including studies that model mediating and moderating relationships among latent constructs.