The Analysis of Attitude Items

Abstract

Surveys often include one or more sets of questions designed to measure a certain attitude. The usual strategy is to ask many such questions and to select the best ones for further analysis or scaling. Unfortunately, the common methods for selecting the best items are based almost exclusively on the correlations among the items themselves. Such methods cannot be counted on to select substantively meaningful and stable subsets of items, because they ignore the relationships of the items to other theoretically relevant variables. This paper suggests a few techniques that researchers might use in order to take such relationships into account and, consequently, to create more useful and meaningful scales of attitude items.