Often contrasted with Cronbach’s coefficient alpha, thus creating a lovely “alpha and omega” of internal reliability/consistency.
Details #
Cronbach’s alpha and McDonald’s omega are both measures of the internal reliability/consistency of a multi-item measure such as any self-report questionnaire (but also applicable to multi-item ratings and interview coding systems). Both measures (and others) try to give us an indication of how much of the variance in a set of responses on measure can be regarded as “true variance” or “signal” as opposed to “error variance” or “noise”. There are no perfect ways to do this but some approaches are certainly better than others. Cronbach’s coefficient alpha was first described by Lee Cronbach in 1951 though he is very careful to observe that he wasn’t inventing it, nor the first person to propose the same index, however, he explored it very carefully and the paper, and the coefficient, became the dominant measure of internal reliability and probably remains that. McDonald’s omega was first described in 1999 and is actually a collection of coefficients with different assumptions. There are arguments that omega is a “better” measure than alpha on the grounds that alpha is not a good indicator that a measure is unidimensional and that it makes quite strong assumptions about the nature of the responding. Both are true and it’s true that (the appropriate) omega goes some way to avoid these issues but it also has issues too complicated for us here. I tend to report both Cronbach alpha (with a bootstrap confidence interval) and the appropriate omega (but not with a CI as I’m not convinced that we have a robust way to get a CI for it).
Try also #
Cronbach’s coefficient alpha
Internal reliability
Psychometrics
Reliability
Chapters #
Not mentioned directly there but chapter covers the general area of psychometrics.
Online resources #
None currently.
Dates #
First created 5.xii.23.
