Confidence Interval Methods for Coefficient Alpha on the Basis of Discrete, Ordinal Response Items: Which One, if Any, is the Best?

Abstract

Abstract In this study, the authors aimed to examine 8 of the different methods for computing confidence intervals around alpha that have been proposed to determine which of these, if any, is the most accurate and precise. Monte Carlo methods were used to simulate samples under known and controlled population conditions wherein the underlying item distribution is nonnormal and when the items' responses are those of rating scales rather than dichotomous items. Overall, one can conclude that, despite concerns expressed over the use of Fisher's method for coefficient alpha, in general, it actually outperformed the other methods. Larger sample sizes and larger coefficient alphas also resulted in better band coverage, whereas smaller number of items resulted in poorer band coverage. Keywords: coefficient alphaconfidence intervalsFisher's z transformationsrating scalesreliability