Abstract
A coefficient is described which can be used to obtain rapid reliability estimates for scales derived from either factor or cluster analysis. Some affinities with Likert and Guttman scaling are noted. But the total approach is superior to either of these techniques, and to factor or cluster analysis used alone, insofar as it enables the researcher not only to explore the dimensionality of an attitude domain, but also to decide, at the same time, on the number of items required in order to measure each main dimension at an appropriate level of reliability. The degree of reliability to be aimed at in sample survey work is considered, together with associated problems of scale validity, item selection and scoring, and assumptions about the level of measurement.
Get full access to this article
View all access options for this article.
