Methodological Issues in Attitude Scale Construction

Abstract

Summary Despite much reliance on summated rating scales, little research has examined the consequences of relying on alternative item analysis strategies. All tend to be treated as though they were comparable even though the statistical methods on which each depends are very different. This analysis, based on data obtained from 3334 respondents who returned mailed questionnaires during the course of a large public opinion survey, addresses this issue by comparing the products of three traditional methods of item selection: t tests, item-to-scale correlations, and factor analysis. The findings demonstrate that the content of attitude measures is significantly affected by the item analysis method employed. Thus, these three methods are far from comparable. At the same time, an exploratory assessment of how well scales based on different item analysis methods perform as predictor variables did not identify major differences. Nevertheless, the results should alert researchers to the fact that item selection m...