Abstract

Ordinal Likert-type indicators are ubiquitous in behavioral and social science measurement, yet applying estimators designed for continuous normal variables can bias parameters and inflate misfit under skewed category use and floor/ceiling effects. This study benchmarks traditional and robust ordinal CFA estimators (WLS, WLSMV, DWLS) using simulated datasets (n = 250, 500, 1000) and an empirical Malaysian Green Consumption dataset (N = 375). All CFA models were estimated on polychoric correlation matrices and evaluated using CFI, TLI, RMSEA, and SRMR. Estimator stability was assessed via nonparametric bootstrapping on the real dataset (B = 500), summarizing convergence rates and average 95% confidence-interval widths for standardized loadings. In addition, machine learning (RF, GBM, SVM) and deep learning (DNN, CNN, RNN) models were evaluated for outcome prediction using 5-fold cross-validation (R², RMSE, MAE). Results show that robust estimators consistently improve fit and stability relative to WLS in small samples (e.g., at n = 250, WLSMV achieved CFI ≈ 0.97 and RMSEA ≈ 0.05 versus WLS CFI ≈ 0.94 and RMSEA ≈ 0.08), and reduce uncertainty in loadings (mean CI width ≈ 0.14–0.15 versus 0.20 for WLS) with near-perfect bootstrap convergence. For prediction, nonlinear learners perform best, with DNN (R² ≈ 0.35) and GBM (R² ≈ 0.33) outperforming other baselines. Overall, the findings provide practical guidance for estimator choice, stability reporting, and predictive validation when analyzing ordinal data.

References

• A. Agresti, Analysis of Ordinal Categorical Data, 2nd ed. Wiley, 2010.
• D. B. Flora and P. J. Curran, “An empirical evaluation of alternative methods of estimation for confirmatory factor analysis with ordinal data,” Psychological Methods, vol. 9, no. 4, pp. 466–491, 2004.
• F. P. Holgado-Tello, S. Chacón-Moscoso, I. Barbero-García, and E. Vila-Abad, “Polychoric versus Pearson correlations in exploratory and confirmatory factor analysis of ordinal variables,” Quality & Quantity, vol. 44, no. 1, pp. 153–166, 2010.
• U. Olsson, “Maximum likelihood estimation of the polychoric correlation coefficient,” Psychometrika, vol. 44, no. 4, pp. 443–460, 1979.
• K. G. Jöreskog, “On the estimation of polychoric correlations and their asymptotic covariance matrix,” Psychometrika, vol. 59, no. 3, pp. 381–389, 1994.
• C. G. Forero, A. Maydeu-Olivares, and D. Gallardo-Pujol, “Factor analysis with ordinal indicators: A Monte Carlo study comparing DWLS and ULS estimation,” Structural Equation Modeling, vol. 16, no. 4, pp. 625–641, 2009.
• C.-H. Li, “Confirmatory factor analysis with ordinal data: Comparing robust maximum likelihood and diagonally weighted least squares,” Behavior Research Methods, vol. 48, no. 3, pp. 936–949, 2016.
• B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap. Chapman & Hall/CRC, 1993.
• C. DiStefano and G. B. Morgan, “A comparison of diagonally weighted least squares robust estimation techniques for ordinal data,” Structural Equation Modeling, vol. 21, no. 3, pp. 425–438, 2014.
• T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning, 2nd ed. Springer, 2009.
• L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001.
• T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proc. KDD, pp. 785–794, 2016.
• D. P. Kingma and M. Welling, “Auto-encoding variational Bayes,” in Proc. ICLR, 2014.
• D. L. Bandalos, “Relative performance of categorical diagonally weighted least squares and robust maximum likelihood estimation,” Structural Equation Modeling, vol. 21, no. 1, pp. 102–116, 2014.
• L.-T. Hu and P. M. Bentler, “Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives,” Structural Equation Modeling, vol. 6, no. 1, pp. 1–55, 1999.
• Y. Rosseel, “lavaan: An R package for structural equation modeling,” Journal of Statistical Software, vol. 48, no. 2, pp. 1–36, 2012.