Brian Evans


On the Difference Between Reliability of Measurement and Precision of Survey Instruments


Despite the importance of assessing the reliability of evaluative measures, there is a confusing array of conceptual scheme, and coefficients for calculating "the reliability" of an item or scale. In particular, reliability may be conceived of and estimated from a true-score model or a sampling precision perspective. The former model is associated with such estimates as parallel or alternate forms reliability, split-half reliability, and coefficient alpha, the latter with standard error, coefficient of variation, and confidence intervals for observed scores. This review clarifies the distinction between two models. The basic theoretical models for each approach are developed and illustrated using data from the author's work on measuring organizational climate. As a result, evaluators should be better able to judge the meaning of the reliability information provided in reports, and to calculate reliability in situations requiring some assessment of the quality of their data.