The goal for any social scientist conducting a survey is to develop a rating on some attitude, value or opinion — a summated rating scale. Aimed at helping researchers construct more effective scales, Spector shows how to determine the number of items necessary, the appropriate amount of response categories and the most productive wording of items. The author discusses how to sort good items from bad (including item-remainder coefficients and Cronbach's alpha) and how to validate a scale, including dimensional validity from factor analysis. User-friendly, the book concludes with a step-by-step account of how to develop a summated rating scale based on classical test theory.
The most difficult part of scale development is validation—that is, interpreting what the scale scores represent. If it is internally consistent, the scale certainly measures something, but determining the nature of that something is a complex problem. Part of the difficulty is that validation can only occur within ...