Abstract

Partial Least Squares Structural Equation Modeling (PLS-SEM) examines complex relationships between composites (linear combinations of observed variables) guided by prior theoretical knowledge. While it has been significantly influential in fields like psychometrics, education, environmental studies, and business research, its potential in medical research remains largely unexplored. PLS-SEM uses iterative alternating Ordinary Least Squares (OLS) algorithms and focuses on the conditional means of outcome variable distributions. However, models focused solely on conditional expectations can be ineffective, especially with skewed response variables, heteroscedasticity, and outliers. To address these issues, we propose a Quantile Composite-based Path Modeling (QC-PM) approach [2], which complements PLS-SEM by examining the entire distribution of outcomes. We emphasized the utility of PLS-SEM in health science and medical research through a real data application, which makes it possible to highlight the capabilities of QC-PM to uncover heterogeneity in the relationships between variables and its advantages as a complementary approach to PLS-SEM.
We recommend exploring both PLS-SEM and QC-PM to address complex research questions, advocating for their broader adoption in medical research.