Replication Code for "From Structural Equation Modelling to Double Machine Learning: Robustness Analysis for Survey-Based Research"

Summary

This repository contains the complete replication package (code, dataset, and outputs) for the paper: "From Structural Equation Modelling to Double Machine Learning: Robustness Analysis for Survey-Based Research", which is publicly available from the arXiv.org open-access archive:

• https://arxiv.org/abs/2607.00512

The application of Structural Equation Modelling (SEM) to Double Machine Learning (DML) for robustness tests is demonstrated using survey data obtained for a FinTech research project.

This workbook is available in Google Colab for viewing and download:

• https://colab.research.google.com/github/usq0/sem_dml_workbook/blob/main/SEM_OLS_DML_Robustness_Workbook_v7_1_public_release.ipynb

Brief Abstract

Structural Equation Modelling (SEM) is widely used in survey-based research to assess latent constructs, but path significance can be sensitive to model specification. This study introduces a staged robustness framework that connects SEM, Ordinary Least Squares (OLS) regression, and Double Machine Learning (DML) to verify the stability of structural relationships. Using SEM-derived construct scores, the framework applies OLS as a benchmark and DML-style residualisation to control flexibly for observed confounders using machine learning learners (Random Forest, Gradient Boosting, and Support Vector Machines). The workflow is demonstrated using a FinTech Digital Customer Intimacy survey model, highlighting which relationships remain robust across multiple estimation frameworks.

Contents of this Deposit

• "SEM_OLS_DML_Robustness_Workbook_v7_1.ipynb": The primary Google Colab/Jupyter notebook containing data preprocessing, model execution, and robustness visualizations.
• "FinTech_data.csv": The survey dataset used for the empirical analysis.
• "sem_ols_dml_experimental_results_v7_0_2.zip": Automated CSV output files containing the empirical results, coefficients, and robustness checks generated by the code.

Requirements & Setup

The computational notebook is fully compatible with Google Colab. To run it locally, you will need a standard Python 3.x environment with machine learning libraries installed (such as `DoubleML`, `scikit-learn`, `statsmodels`, `pandas`, and `numpy`).

Instructions

1. Download all files into the same working directory (or upload them to your Google Drive).
2. Open the `SEM_OLS_DML_Robustness_Workbook_v7_1.ipynb` notebook in Google Colab or Jupyter.
3. Run the cells sequentially to regenerate the provided CSV results.