Skip to content
Omnikalc

Composite reliability and AVE calculator

Enter the standardized factor loadings of five items to get the composite reliability, the average variance extracted, its square root for the Fornell-Larcker test and the alpha implied by the loadings.

Updated on
Your inputs

Value out of range

Value out of range

Value out of range

Value out of range

Value out of range

Results

Composite reliability (omega)

0.848

Average variance extracted (AVE)

0.529

Square root of the AVE
0.727
Standardized alpha implied by the loadings
0.846
Total error variance of the items
2.354
Mean loading
0.724
Passes the usual checks (CR ≥ 0.70 and AVE ≥ 0.50: 1 = yes)
1

The composite reliability and AVE calculator gives the composite reliability (McDonald’s omega), the average variance extracted, its square root and the alpha implied by the loadings, for a scale of five items, from their standardized factor loadings.

How the calculation works

  • Enter the standardized factor loadings of the five items from a factor analysis.
  • The calculator computes the error variance of each item as 1 minus the squared loading.
  • Read the composite reliability, the AVE and the checks.

Example

Five items with loadings of 0.82, 0.78, 0.72, 0.68 and 0.62, a mean of 0.724, have a total error variance of 2.354. The composite reliability is 0.848 and the average variance extracted 0.529, whose square root is 0.727. The alpha implied by the loadings is 0.846. The check on the usual thresholds gives 1.

Why it matters

The calculator helps researchers in psychology, marketing, education and management who validate a questionnaire scale with a confirmatory factor analysis or a structural equation model, and need to report the reliability and the convergent validity of each construct.

Good to know

Use the standardized loadings of a one-factor model or of the items of one construct in a larger model, and make sure that the model fits. The usual thresholds are 0.70 for the composite reliability and 0.50 for the AVE, but they are guidelines. An AVE under 0.50 with a composite reliability above 0.60 is sometimes accepted. Loadings above 0.95 may signal redundant items, and cross-loadings need a separate check. Report the loadings themselves in a table, along with the sample size and the estimation method, so that readers can recompute the indices and judge the fit of the measurement model.

The Cronbach alpha calculator gives the classical reliability, the Kuder-Richardson calculator treats right-wrong items, the standard error of measurement calculator turns a reliability into a score band and the Spearman-Brown calculator predicts the effect of test length.

Frequently asked questions

What is composite reliability?

Composite reliability, or McDonald’s omega in the one-factor case, estimates the reliability of a scale from the factor loadings of its items. It is CR = (Σλ)² / ((Σλ)² + Σθ), where λ are the loadings and θ = 1 − λ² the error variances of standardized items. Unlike Cronbach’s alpha, it does not assume that all the loadings are equal.

What is the average variance extracted?

The AVE is the mean of the squared loadings, the share of the variance of the items that the construct explains on average: AVE = Σλ² / k. A value of 0.50 or more means that the construct explains more than half of the variance of its items, and is the usual criterion of convergent validity of Fornell and Larcker.

How do I test discriminant validity?

In the Fornell-Larcker approach, the square root of the AVE of each construct must be larger than its correlations with the other constructs. The calculator gives the square root of the AVE for this comparison. Other methods, like the heterotrait-monotrait ratio, are now often preferred.

Why differ from Cronbach’s alpha?

Alpha is exact only when all the items have the same loading. When the loadings differ, as in most scales, alpha underestimates the reliability. The alpha implied by the loadings is shown so that you can compare. The composite reliability is usually higher, by a few hundredths.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.