You are on page 1of 18

How to estimate composite reliability/McDonald’s omega for a one-factor CFA

model in AMOS using a ‘phantom variable’ approach

Mike Crowson, Ph.D.


University of Oklahoma
December 31, 2020
Re-analysis of data associated with:

Leal-Soto F, Ferrer-Urbina R (2017). Three-factor structure for Epistemic Belief Inventory: A cross-
validation study. PLoS ONE 12 (3): e0173295. doi:10.1371/journal.pone.0173295
The basic one-factor CFA model being tested in this
demonstration
Unstandardized estimates
Standardized estimates
Fit information associated with the CFA
model
(∑ 𝜆 𝑖 )
2

𝜔= Equation 1: No correlated errors If you are not using the


( ∑ 𝜆𝑖 ) + ∑ 𝜃𝑖𝑖
2
phantom variable approach,
then you would need to use
these formulas for computing
composite reliability /
( ∑ 𝜆𝑖 )
2
McDonald’s omega
𝜔= Equation 2: With correlated errors
( ∑ 𝜆𝑖 ) + ∑ 𝜃𝑖𝑖 + 2 ∑ 𝜃𝑖𝑗
2

included in model

𝜙 11 ( ∑ 𝜆𝑖 )
2

𝜔= Equation 1 if factor variance () is not fixed to 1


𝜙11 ( ∑ 𝜆𝑖 ) + ∑ 𝜃 𝑖𝑖
2

𝜙11 ( ∑ 𝜆 𝑖 )
2

𝜔= Equation 2 if factor variance () is not fixed to 1


𝜙11 ( ∑ 𝜆𝑖 ) + ∑ 𝜃 𝑖𝑖 +2 ∑ 𝜃𝑖𝑗
2
Model specification to compute composite reliability / McDonald’s omega

Latent factor with variance fixed to 1.0

A phantom variable representing the


summation or composite of the
indicator variables.
Square the model-implied correlation between the phantom (composite) variable and the latent factor
to obtain McDonald’s omega.

For this analysis, McDonald’s omega is (.85455)2 = .730.

Compare result with that shown in Appendix A.


It is possible to generate a bootstrap interval estimate of reliability in AMOS if you do not have missing data on
your indicator variables.
Square the Lower Bound and Upper Bound estimates of the model-implied correlation between the
phantom (composite) variable and the latent factor to obtain an interval estimate for McDonald’s omega.
The Lower Bound estimate is (.84197)2 = .709. The Upper Bound estimate is (.86606)2 = .750.
The bootstrap 95% confidence interval estimate for McDonald’s omega is (.709,.750).
Alternative specification, where
an error term is added to the
phantom variable, but the
variance is fixed to 0.

This is the approach shown by Hayes and Coutts (2020).


Strong suggestion:

If you have a unidimensional scale containing both positively- and negatively-worded items, I encourage you
to consider reverse coding the negatively worded items prior to attempting to compute composite reliability
via the phantom variable method (or using the formulas I include in this PowerPoint). Failure to do so can
result in substantial underestimation or even nonsensical estimates of reliability since negative factor
loadings will be subtracted from – rather than added to – the summed loadings in the numerator and
denominator in the reliability formula.

If your scale contains items that are worded in only one direction and the factor loadings are all positive,
then there is no need to recode prior to computing reliability.
References and further reading

Bagozzi, R.P., & Yi, Y. (2012). Specification, evaluation, and interpretation of structural equation models. Journal of the
Academy of Marketing Science, 40, 8-34.

Brown, T.A. (2006). Confirmatory factor analysis for applied research. New York: The Guilford Press.

Cho, E., & Kim, S. (2015). Cronbach’s alpha: well known but poorly understood. Organizational Research Methods, 18, 207-
230.

Fornell, C., & Larker, D.F. (1981). Evaluating structural equation models with unobservable variables and measurement
error. Journal of Marketing Research, 18, 39-50.

**Graham, J. (2006). Congeric and (essentially) tau-equivalent estimates of score reliability. Educational and Psychological
Measurement, 66, 930-944.

**Hayes, A.F., & Coutts, J.J. (2020). Use omega rather than Cronbach’s alpha for estimating reliability. But… Communication
Methods and Measures, 14, 1-24.

Kline, R. B. (2016). Principles and practice of structural equation modeling (4th ed.). New York: The Guilford Press.

**Articles that explicitly discuss or demonstrate estimation of reliability using AMOS.


References and further reading (cont’d)

McNeish, D. (2018). Thanks coefficient alpha: we’ll take it from here. Psychological Methods, 23, 412-433.

Padilla, M.A., & Divers, J. (2016). A comparison of composite reliability estimators: coefficient omega, confidence intervals
confidence intervals in the current literature. Educational and Psychological Measurement, 76, 436-453.

Raykov, T. (1997). Estimation of composite reliability for congeneric measures. Applied Psychological Measurement, 21,
173-184.

Trizano-Hermosilla, I., & Alvarado, J.M. (2016). Best alternatives to Cronbach’s alpha reliability in realistic conditions:
Congeneric and asymmetrical measurements. Frontiers in Psychology, 7:769. doi: 10.3389/fpsyg.2016.00769

Note: Some authors refer explicitly to McDonald’s omega, whereas others (particularly coming from an SEM standpoint)
refer to composite reliability. Some refer to rho or ‘factor-rho coefficient’(see Kline, 2016). Nevertheless, the effect of
the phantom variable approach is to compute the reliability estimates congruent with those computed using
formulas for computing McDonald’s omega.
Appendix: Comparison with results generated using
Hayes’ OMEGA macro in SPSS
Results on left include unstandardized factor loadings based on OMEGA macro (in SPSS). Results on right are unstandardized
loadings generated for our analysis in AMOS. Omega for both is .73. [Coefficients were generated in both using ML
estimation. They are the same (with only slight differences due to rounding).

You might also like