COMPARING COMPETING STRUCTURAL MODELS IN CB-SEM: AN ILLUSTRATIVE APPLICATION FOR MANAGEMENT AND BEHAVIORAL RESEARCH

Comparing Competing Structural Models in CB-SEM

Abstract

Structural equation modeling (SEM) is an advanced statistical technique used to evaluate complex relationships among latent variables. Researchers in behavioral and management sciences frequently apply structural models right after validating the measurement model, often without considering other plausible alternative (competing) models, perspectives, or explanations. Consequently, competing structural models are rarely evaluated or reported. This problem is compounded by practices such as selectively using fit indices, adopting more complex models for increased explanatory power, and relying on software-generated modification indices without a strong theoretical foundation. To improve robustness, both the hypothesized model and alternative (competing) models should be compared using fit indices and descriptive criteria for model selection, such as the Akaike Information Criterion (AIC). This paper shows how to compare a hypothesized model with competing models—both nested and non-nested—using covariance-based SEM techniques (CB-SEM; Analysis of Moment Structures [AMOS] and Linear Structural Relations [LISREL]). Nested models were compared using the χ2 difference test, while AIC was employed for non-nested models. A dataset from a published study (N=282), based on the Technology Acceptance Model, was analyzed. Results from AMOS indicated that the competing models met acceptable fit criteria just as the primary structural model did. LISREL output provided higher incremental fit indices and path coefficients than AMOS, yet both identified the same preferred model, with similar significance patterns and consistent theoretical interpretations. The study found that the initial structural model achieved better fit statistics than other competing models, whether nested or non-nested. It emphasizes the importance of model comparisons and highlights the value of testing alternative theoretical models. The paper offers a comprehensive step-by-step methodological guide for researchers to conceptualize, execute, and report these analyses. Finally, it discusses implications and recommendations for future research.