Introduction to Item Response Theory
The discussion in the last few chapters evolved mainly within the framework of classical test theory (CTT). An important limitation of CTT is that it does not place routinely in the center of its concerns how individuals at different levels of the construct studied (ability, trait, attribute) perform on the components, or items, of an instrument aimed at measuring that underlying latent dimension. This information is focused on by an alternative approach to test development, item response theory (IRT), at times also referred to as latent trait theory (cf., e.g., McDonald, 1999). Informally, a main assumption in IRT is that the responses on items of a test under consideration (and consequently overall test performance) can be accounted for by one or more latent abilities or constructs, which are usually much fewer in number than the items. Most current applications of IRT assume that there is a single latent trait behind the responses to items on a given instrument. Although much progress has been made in the past 20 years or so in multidimensional IRT (e.g., Reckase, 2009), the majority of empirical utilizations of IRT still seem to be concerned with single traits. Given this tendency, in addition to the confines of this book, in its remainder we will assume (unless otherwise indicated) that there is a single underlying latent dimension along which individuals differ and that accounts for the relationships among their responses on a given set of components or items comprising a measuring instrument under consideration.