Title: Your coefficient alpha is probably wrong, but which coefficient omega is right?

Abstract: Accurate measurement is fundamental to any science, as it is well known that measurement error leads to imprecise (if not entirely incorrect) results. Reliability quantifies the influence of measurement error on an observed score and is formally defined as the proportion of true score variance relative to total score variance. Despite its popularity, coefficient alpha is an unbiased estimate of this proportion only under strict conditions which are unlikely to hold in any empirical situation. Consequently, many psychometricians have argued that alpha should be abandoned. An alternative reliability estimate, known as coefficient omega, relaxes the strict assumptions of alpha and can provide an accurate estimate of the proportion of total test score variance that is due to an unobservable construct. Yet, there are several different forms of omega which differ according to the underlying factor structure of the total score components (e.g., item responses), among other considerations, and choosing among them is challenging. In this talk, I will describe how to calculate different forms of omega using canned functions available in R and provide advice for selecting and appropriately interpreting these reliability estimates.