EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A hybrid factor analysis framework for ordinal data integrating CFA, ML, and DL methods

Abuelgasim Osman, Zahayu Md Yusof

Frontiers in Applied Mathematics and Statistics · 22.06.2026 · 12

DOI: 10.3389/fams.2026.1830237 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Applied Mathematics and Statistics, online erschienen am 22.06.2026. Verfasst von Abuelgasim Osman und 1 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ordinal Likert-type indicators are pervasive in behavioral and social-science research, yet they violate the continuity and multivariate-normality assumptions that underlie many conventional factor-analytic procedures. Robust ordinal confirmatory factor analysis (CFA) estimators such as weighted least squares mean and variance adjusted (WLSMV) and diagonally weighted least squares (DWLS) reduce parameter bias, but they remain fundamentally linear and may leave threshold-related nonlinearities or item-level interactions in the residual structure. This study proposes a Hybrid Factor Analysis (HFA) framework that integrates robust CFA with machine-learning and deep-learning item signals through an adaptive weighting parameter, ω . The hybrid measurement weight, λ_combined, is constructed as a convex combination of normalized CFA loadings and normalized computational item-importance measures. The framework was evaluated in a Monte Carlo design for a five-factor, 36-indicator ordinal model under sample sizes of n = 250, 500, and 1,000, and then examined empirically using the Green Consumption Behavior dataset from Malaysia ( n = 375). Under highly skewed thresholds and n = 250, HFA reduced loading mean squared error (MSE) from 0.092 to 0.068 relative to standalone WLSMV, a 26.1% improvement. In the empirical application, HFA improved global fit from CFI = 0.942, TLI = 0.938, and RMSEA = 0.068 under standalone WLSMV to CFI = 0.965, TLI = 0.959, and RMSEA = 0.048, while reducing the average bootstrap confidence-interval width from 0.124 to 0.089. The hybrid model also improved predictive performance over the linear CFA baseline (R 2 = 0.32 vs. 0.28) while preserving confirmatory interpretability. These findings position HFA as a next-generation framework for ordinal psychometrics in which theory-constrained measurement and computational learning are treated as complementary rather than competing sources of information.

Zitieren

APA

Osman, A., & Yusof, Z. M. (2026). A hybrid factor analysis framework for ordinal data integrating CFA, ML, and DL methods. Frontiers in Applied Mathematics and Statistics, 12. https://doi.org/10.3389/fams.2026.1830237

Vancouver

Osman A, Yusof ZM. A hybrid factor analysis framework for ordinal data integrating CFA, ML, and DL methods. Frontiers in Applied Mathematics and Statistics. 2026;12. doi:10.3389/fams.2026.1830237

BibTeX
@article{Osman2026,
  author = {Osman, Abuelgasim and Yusof, Zahayu Md},
  title = {A hybrid factor analysis framework for ordinal data integrating CFA, ML, and DL methods},
  journal = {Frontiers in Applied Mathematics and Statistics},
  year = {2026},
  volume = {12},
  doi = {10.3389/fams.2026.1830237}
}
Kostenlos starten →