Reliability

Abstract

Reliability of test scores is discussed for four different approaches: classical test theory, generalizability theory, factor analysis, and item response theory. For each approach, we discuss the test design, the model on which the approach is based, methods for estimating reliability, measurement precision for assessing individual test performance, and we provide comments. The four approaches are compared, and merits and drawbacks are discussed. We emphasize the difference between reliability and validity: For each model, validity—the question whether the test covers the attribute of interest well—must be studied separately from the estimation of test-score reliability.