Partial Least Square (PLS) Model Estimator with PATHMOX Approach Using Mediation Variables

Abstract

One of the multivariate statistical analysis techniques that can explain complex relationship patterns involving many variables is Structural Equation Modeling (SEM). Variance-Based SEM (VB-SEM), also known as Partial Least Squares (PLS), is a more flexible method in SEM because it is able to overcome various limitations, such as the scale of different variable indicators, limited sample counts, and data distribution assumptions. PATHMOX is a special strategy in PLS-SEM that is used to segment structural model results to overcome the observed heterogeneity. This study aims to build a structural equation model (SEM-PLS) and segment the results of structural equation models using the PATHMOX approach. The PATHMOX algorithm is carried out by comparing structural models between segments through the F-global test. The path coefficient is estimated with OLS as in multiple linear regression analysis between latent variables  and .