Comparison of Covariance-Based Structural Equation Model and Partial Least Squares Equality Models

Abstract

Duygu Vargor and Tuncay Ogretmen

The aim of this research is to compare the differences between the objectives, distribution assumptions, sample sizes, parameters, fit indices, and measurement models of the covariance-based structural equation model (CB-SEM) and the partial and consistent partial least squares structural equation models (PLS-SEM and PLSc-SEM) to contribute to future studies. Data from Turkey's Information and Communication (ICT) scale of the Program for International Student Assessment (PISA) for the year 2018 was used. Exploratory factor analysis (EFA) was initially conducted on the data from a sample of 5963 individuals, followed by confirmatory factor Analysis (CFA) using CB-SEM, PLS-SEM, and PLSc-SEM. CFA was performed by obtaining normal and non-normal distributions from the same sample data. The structure validity and reliability, goodness-of-fit indices, item parameters, and latent variable parameters obtained using CB-SEM, PLS-SEM, and PLSc-SEM were compared. The CB-SEM model fit indices provide a better method for explaining how well a hypothetical model fits the experimental data. PLS-SEM and PLSc-SEM, on the other hand, have sufficient reliability and validity parameters for the weight of the items, while the confidence intervals, estimations, and variances of the items are insufficient. This study concludes that it is not appropriate to claim that PLS-SEM is a preferred method when the sample size is small, and the data distributions are non-normal. It is essential for the observed data to be consistent with the hypothesis and theory; otherwise, the analysis results may lead to errors and misconceptions.

PDF