Robustness Framework Connects SEM, OLS, and DML for Surveys

Survey-based research, particularly in business and information systems, frequently employs Structural Equation Modeling (SEM) to analyze latent constructs and theoretical relationships. However, the significance of SEM path coefficients can be sensitive to the specific model specification, raising questions about the stability of findings under alternative analytical approaches. To address this, a new staged robustness analysis framework has been developed, integrating SEM, Ordinary Least Squares (OLS) regression, and Double Machine Learning (DML). The framework begins with SEM to refine measurement structures and establish a baseline model, retaining the full theoretical path system for subsequent robustness checks. OLS regression is then applied to SEM-derived construct scores, serving as a transparent benchmark. Finally, DML-style residualization is used to rigorously test whether focal relationships remain stable after flexible, machine-learning-based adjustments for observed control variables. The study includes learner-sensitivity checks with Random Forest, Gradient Boosting, and Support Vector Machine, along with reverse-direction diagnostics. Demonstrated with a FinTech Digital Customer Intimacy survey, the framework identifies relationships that are consistently stable across all methods versus those requiring cautious interpretation, providing a reproducible template for researchers.