Psychometrics with R: A review of CRAN packages for item response theory

Thomas Rusch, Patrick Mair, Reinhold Hatzinger

Publication: Working/Discussion PaperWU Working Paper

92 Downloads (Pure)

Abstract

In this paper we review the current state of R packages for Item Response Theory (IRT). We group the available packages based on their purpose and provide an overview of each package's main functionality. Each of the packages we describe has a peer-reviewed publication associated with it. We also provide a tutorial analysis of data from the 1990 Workplace Industrial Relation Survey to show how the breadth and exibility of IRT packages in R can be leveraged to conduct even challenging item analyses with versatility and ease. These items relate to the type of consultations that are carried out in a firm when major changes are implemented. We first use unidimensional IRT models just to discover that they fit do not fit well. We then use nonparametric IRT to explore the possible causes for the scaling problem. Based on the results from the exploration, we finally use a two-dimensional model on a subset of the original items to achieve a good fit with a sensible interpretation, namely that there are two types of consultations a firm may engage in: consultations with workers/representatives from the firm and with official union representatives. The different items relate mostly to one of these dimensions and firms can be scaled well along these two dimensions.
Original languageEnglish
Publication statusPublished - 2013

Publication series

NameDiscussion Paper Series / Center for Empirical Research Methods
No.2013/2

WU Working Paper Series

  • Discussion Paper Series / Center for Empirical Research Methods

Cite this