Hierarchical (second-order) versus bifactor CFA

Description

Tests whether a general factor plus specific factors explains a multidimensional scale, and whether a total score, subscale scores or both are defensible.

Many UX questionnaires report both a total and subscales without testing whether either is justified. In a second-order model, first-order factors (subscales) load on a general factor. In a bifactor model each item loads directly on a general factor and on one specific factor, with all factors orthogonal. The second-order model is nested within the bifactor model, so they can be compared with a chi-square difference test, and the Schmid-Leiman transformation converts one to the other. Omega-hierarchical estimates how much total-score variance the general factor explains, indicating whether a total score is meaningful; subscale omegas show whether specific factors add reliable variance beyond it. Bifactor models tend to fit well for purely statistical reasons, so the choice should be stated a priori and judged on residuals and omega values, not on fit alone.

Attributes

Loss type
Siloed
Stage
Evaluation
Effort
Days

Steps

  1. State a priori whether a general factor is expected and how the specific factors relate to it.
  2. Fit the first-order correlated-factors CFA as a baseline.
  3. Fit the second-order and bifactor models.
  4. Compare nested models with chi-square difference tests and inspect residual correlations.
  5. Compute omega-hierarchical for the total score and omega-hierarchical-subscale values for each specific factor.
  6. Decide whether to report a total, subscales or both, and document the reasoning.

Sources

  • Principles and Practice of SEM (Kline, 2023)
  • A Beginner's Guide to Structural Equation Modeling (Schumacker & Lomax, 2004)
  • Latent Variable Models (Loehlin & Beaujean, 2017)
  • The SAGE Handbook of Quantitative Methods in Psychology (Millsap & Maydeu-Olivares, 2009)
  • R for Marketing Research and Analytics (Chapman & Feit, 2019)

Also known as

  • Bifactor model
  • Second-order factor model
  • Higher-order CFA
  • Schmid-Leiman transformation
  • Omega-hierarchical

Categories

  • Analytics & Behaviour
  • Survey & Psychometrics