Generalizability Theory: Conceptual Framework
Hoi K. Suen · 2012
W hether a paper-and-pencil test, a rating scale, or a self-report questionnaire is used, a test is most frequently made up of a number of items (although there can be single-item rating scales). Conceptually, if we know the quality of each item in a test, we should be able to deduce the quality of the total test score. Minimally, knowledge of the quality of each item should provide indications of the reliability of the total score. Unfortunately, within classical and generalizability theory, there are no established procedures through which one can deduce the reliability of the total score from known characteristics of individual items. However, knowledge of the characteristics of individual items can guide us in the improvement of the test, and thus maximize the ultimate reliability of the total score.