Integration of ISSM-TAM Models for Sustainable E-Learning from a Bayesian Structural Approach

The application of the Bayesian approach emerges as a plausible alternative in multivariate statistical analysis within structural equation models (SEM), being more precise, reliable, and robust for parameter estimation, especially in situations where traditional methods may not be adequate due to sample size restrictions, non-compliance with multivariate normality, or high theoretical complexity. The study aimed to determine the relevance of the proposed ISSM-TAM e-learning model on the academic performance of students in the 2023-II cycle at the National Agrarian University La Molina (UNALM). The methodology was quantitative, basic, explanatory, non-experimental, and cross-sectional. A sample of 767 undergraduate students was collected. A questionnaire with 10 latent constructs, comprising 40 items measured on a 1-5 ordinal Likert scale, was used. Various first and second-generation multivariate statistical techniques were applied: a variance-based structural equation model (PLS-SEM), covariance-based model (CB-SEM), and Bayesian model (B-SEM) with non-informative prior distributions; uniform and reference priors, and a specific informative prior distribution. The obtained data were analyzed exploratorily, descriptively, and inferentially, validating the measurement and structural models with both frequentist and Bayesian approaches. It was found that the Bayesian model with informative priors achieved acceptable values in goodness-of-fit indices compared to the others, and it explained the structural relationships between the main endogenous latent variables of the proposed hybrid model. Received: 5 January 2025 / Accepted: 29 August 2025 / Published: 05 September 2025