A Network approach to deal with the problem of examinee choice under item response theory

Abstract

In a typical questionnaire testing situation, the examinees are not allowed to choose which items they would rather answer. The main reason is a technical issue in obtaining satisfactory statistical estimates of examinees' abilities and items' difficulties. This paper introduces a new Item Response Theory (IRT) model that incorporates information from a novel representation of the questionnaire data, using network analysis. The questionnaire data set is coded as layers, the items are coded as nodes and the selected items are connected by edges. The new proposed Item Response Theory (IRT) model incorporates networkinformation using Bayesian estimation. Several simulation studies in which examinees are allowed to select a subset of items were performed. Results show substantial improvements in the parameters' recovery over the standard model. To the best of our knowledge, this is the first proposal to obtaining satisfactory IRT statistical estimates in some critical scenarios reported in literature.