Description

The book covers thoroughly all that is needed for a full understanding of reliability in terms of classical theory. Ross E. Traub manages to make clear the difference between reliability as a theoretical construct and methods of estimating it in practice. In a useful chapter, he proceeds to examine factors that may affect the reliability of a test, including time limits, test length, item characteristics, subjectively scored items, and heterogeneity of the population taking the test. . . . The book is a useful reference for those with some understanding of reliability theory and would probably make a reasonable textbook (including copious exercises) for those studying the subject from scratch, provided their mathematical background was adequate. --Ian Schagen in Educational Research How can social scientists assess the reliability of the measures derived from tests and questionnaires? Through a careful and illustrative review of the principles of classical reliability theory, Ross E. Traub explores some general strategies for improving measurement procedures. Beginning with a presentation of random variables and the expected values of a random variable, Reliability for the Social Sciences covers such topics as the definition of reliability as a coefficient and possible uses of a coefficient, the notion of parallel tests so as to make possible the estimation of a reliability coefficient for a set of measurements, what to do when parallel tests are not available, what factors affect the reliability coefficient, and how to estimate the standard error of measurement. Aimed at giving readers a nontechnical treatment of classical reliability theory, the book also includes end-of-chapter exercises, as well as boxes that give more in-depth coverage of major topics or that provide algebraic proofs.