Assessment of the distance between categories in rating scales by using item response model

Abstract

This study aimed to assess the distance between adjacent categories of rating scales. It is common practice to treat ordinal variables as interval-scaled variables in the analysis of rating scales. Strictly speaking, however, ordinal scale data should be treated as such, since there is little reason and assurance that they are equivalent to interval variables. In view of this practice, this study proposes a method to assess the interval of rating scales, and analyzes empirical data in order to examine the results obtained by the method. This method is based upon the generalized partial credit model which is one of item response theory (IRT) models. The experiment was carried out on two data sets that differed only on the verbal phrasing of the rating. Main results of the study were: 1) the difference in item content (positive or negative) affects the width of a neutral category; and 2) the distance between categories differs significantly reflecting the difference in verbal phrasing.