Abstract
This article is a pedagogical piece on coefficient alpha (alpha) and its uses. The classical approach to test reliability is explained. Test-retest, alternative-forms, and internal-consistency methods of approximating test reliability are described, equations are derived for each method, and alpha is shown to be a lower-bound internal-consistency approximation to test reliability. Emphasis is placed on the effects of violations of model assumptions on reliability estimation. The classical models are conceptualized as structural equation models and are displayed in path diagrams. Special emphasis is placed on the failure of alpha to meet certain basic criteria as an index of test homogeneity.