McDonald's Hierarchical Omega (ωh)
Also known as: omega hierarchical, omega-h, bifactor omega, composite score validity coefficient, McDonald's Omega Hiyerarşik (ωh) — Kompozit Puan Geçerliliği
McDonald's hierarchical omega (ωh) is a coefficient derived from a bifactor confirmatory factor model that quantifies what proportion of total-score variance is attributable to a single general factor rather than to group-specific factors or item-level error. Introduced by Roderick P. McDonald (1999) and elaborated for bifactor applications by Reise and colleagues (2013) and Rodriguez and colleagues (2016), it is the primary index used in psychometrics to evaluate whether a composite total score is a defensible summary of a multidimensional scale.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
McDonald's ωh is appropriate when you have a multidimensional scale but wish to report and interpret a single total score, and you need to justify that the composite meaningfully reflects one dominant construct. The method requires at least 200 participants and item responses at ordinal or continuous measurement level. A bifactor model must be theoretically motivated — typically the scale must contain at least two distinguishable group factors each with at least three items — and the model must achieve adequate fit before ωh is interpreted. Do not use ωh as a general reliability substitute for Cronbach's alpha in strictly unidimensional scales; in that context ωt (total omega) or alpha are more appropriate.
Strengths & limitations
- Directly quantifies how much of composite-score variance reflects the general factor, providing evidence that a total score is substantively meaningful even when the scale is multidimensional.
- Grounded in a structural model rather than a formula heuristic, so it explicitly accounts for group-factor variance that inflates Cronbach's alpha in multidimensional scales.
- ECV and other bifactor indices (PUC, ωhs) derived from the same model provide a richer diagnostic picture of scale dimensionality than any single-number reliability coefficient.
- Requires fitting a full bifactor CFA model, which demands a minimum sample of about 200 and can converge poorly when group factors have few items or low loadings.
- ωh is sensitive to model misfit; if the bifactor model itself fits poorly, the coefficient is uninterpretable regardless of its numerical value.
- A high ωh does not guarantee that subscale scores are also reliable — subscale-level omega (ωhs) must be computed separately to evaluate the usefulness of sub-scores.
Frequently asked
How is ωh different from Cronbach's alpha?
Cronbach's alpha estimates the proportion of total-score variance due to all common factors combined — general and group factors alike. If the scale is multidimensional, alpha conflates the general factor with group-factor variance and overstates how purely the total score reflects a single construct. ωh separates them: it isolates only the general-factor contribution. A scale can have high alpha but low ωh, signalling that the apparent reliability is driven by narrow subscale clusters rather than a single dominant trait.
What does ECV add beyond ωh?
ECV (Explained Common Variance) is the proportion of all common variance attributable to the general factor, ignoring error. While ωh speaks to the proportion of total composite variance due to the general factor, ECV focuses on the common-factor space and is less sensitive to item uniqueness. Together, ωh and ECV paint a fuller picture: both should be high for the composite to be a defensible summary of a general trait.
Can I use ωh with a unidimensional scale?
ωh requires a bifactor model with at least one general factor and at least two group factors. If the scale is genuinely unidimensional, there are no group factors to separate out, and ωh reduces to ωt. In that setting, use total omega (ωt) or Cronbach's alpha; ωh adds no information and the bifactor model will likely be under-identified or poorly fitting.
How large a sample is needed?
The bifactor CFA underlying ωh is parameter-intensive. A minimum of 200 participants is commonly cited; more complex models with many group factors may need 300 or more. Below 200, both the bifactor model and the resulting ωh estimate tend to be unstable, and bootstrap confidence intervals become very wide. Always report bootstrap CIs so readers can judge precision.
Sources
- Reise, S. P., Scheines, R., Widaman, K. F. & Haviland, M. G. (2013). Multidimensionality and structural coefficient bias in structural equation modeling: A bifactor perspective. Educational and Psychological Measurement, 73(1), 5–26. DOI: 10.1177/0013164412449831 ↗
- Rodriguez, A., Reise, S. P. & Haviland, M. G. (2016). Evaluating bifactor models: Calculating and interpreting statistical indices. Psychological Methods, 21(2), 137–150. DOI: 10.1037/met0000045 ↗
How to cite this page
ScholarGate. (2026, June 1). McDonald's Hierarchical Omega (ωh). ScholarGate. https://scholargate.app/en/psychometrics/mcdonald-omega
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bifactor ModelPsychometrics↔ compare
- Confirmatory factor analysisPsychometrics↔ compare
- Cronbach's AlphaStatistics↔ compare
- EFAStatistics↔ compare
- SEMStatistics↔ compare