Methodological advances in the validation of inventory items, scales, profiles and interpretations

Abstract

Factor analytic research is still in its infancy-theoretically, methodologically, and interpretively. Definitive research has not yet been done on problems of (a) item design and selection of items suitable for factor analysis; (b) identification of highly significant marker items; (c) tracings of item response distributions across scales, populations, occasions and other relevant variables; (d) determination of the number of items with different loadings necessary to obtain valid scale scores for psychological state measurements; (e) questions about the validity of summating factorially heterogeneous lowly correlated items in arriving at scale scores; (f) the use of clinical judgment to evaluate the possibly differing significances of the same item responses across populations and occasions; i. e., does the same factored item response have the same significance across scales?; (g) how to interpret scales with factorially mixed item compositions; (h) clinical judgment decisions in evaluating and interpreting scale meanings; and (i) the comparison of the results of analyzing data by different scaling methods. This study makes a start toward clarifying these methodological issues by analyzing the results of applying different methods of scale construction that utilized 146,000 item responses made by 730 Ss on the Personal Health Survey, a 200-item inventory designed to measure organ system functioning that contributes to physical and mental health.