Item Response Theory Models for Rating Scale Data

Abstract

Abstract The purpose of this entry is to provide a brief description of several IRT models that can be used to analyze rating scale data. Rating scale data are defined as responses scored in two or more ordered categories. Rating scale data represent the most common formats for collecting information from examinees or raters. The basic idea that motivates the use of IRT models for rating scale data is that the scoring of m + 1 ordered categories with ordered integers (0, 1, …, m ) using the assumption that there are equal intervals between the categories may not be justified. IRT models provide a framework to explicitly examine this assumption, and parameterize the categories without this assumption. The specific IRT models that are described in this entry are direct models (Partial Credit Model, Generalized Partial Credit, and Rating Scale Model) and indirect models (Graded Response Model and Modified Graded Response Model). Graphical procedures are used to illustrate various aspects of these IRT models.