Regularized Structural Equation Modeling for Short-Form Tests in the Presence of Missing Data: A Monte Carlo Simulation

• In Event: Virtual Poster Session
  In Virtual iPresentation Only: Division D - Section 2: Quantitative Methods and Statistical Theory Virtual Poster Session

Abstract

This study aims to investigate the performance of a regularized Structural Equation Modeling (regSEM) in the presence of missing data for the purpose of short-form tests, employing a Monte Carlo simulation. The simulation encompasses 96 different scenarios, varying in missingness mechanisms (MAR, MNAR), missing rates (5%, 30%), missing data techniques (listwise deletion, mean imputation, k-NN, random forest), sample sizes (200, 500, 1000), and analysis methods (MLE, regSEM). The evaluation criteria included convergence rates, fit indices, and variable selection. As results, regSEM demonstrated higher convergence rates compared to MLE and maintained stable fit indices across various conditions. Also, regSEM reduced false positives under specific conditions. Future research topics are suggested.

Authors

• Bitna Park, Korea National University of Education

• Jin Eun Yoo, Korea National University of Education