A comparison between the rating scale model and dual scaling for Likert scales

Abstract

The fundamental requirement of using Likert scales to measure affective behaviors-that all respondents must use the ordered response categories of all scale statements in the same way-is examined. Scaling problems arise when attitude statements within a Likert scale are unfolding preference data, and when the position and spacing of the ordered response categories are interpreted differently across scale statements and among respondents. The assignment of a neutral response category in the center of the response scale is questionable. These problems pertain to issues of the existence of an interval scale, the dimensionality of the trait, and patterns of item response functions. An attempt is made to resolve these problems using two contrasting scaling methods-item response theory modeling and dual scaling. The primary result is that conformity of responses to the item response theory model requirements and targeting of scale statements and their ordered response categories to the respondents in dual scaling are vital for a resolution of these problems. This study establishes the similarity of the two scaling methods. Index terms: dual scaling, items response theory, jackknife technique, Likert scaling, Rasch model, rating scale model.