4: Generalizability Theory

In Chapter 3, we discussed the estimation of reliability indices in terms of internal consistency and correlations between scores on parts (e.g., split-half reliability) or the entire (e.g., test–retest reliability) instrument (e.g., survey, test). In this chapter, we describe evaluating reliability from a very different perspective, and one that is linked more closely to the fundamental equation in CTT than were the other estimates of reliability described previously. This alternative approach to estimating reliability is known as generalizability theory, and as we will see below, it is based upon a framework embedded in a commonly used statistical model, the analysis of variance (ANOVA).