Using Projective Unidimensional Models for Measuring Multidimensional Educational Data

The goal of this project is to establish the foundation of a measurement paradigm that leverages the power unidimensional item response theory (IRT) for handling tests that are multidimensional, even when the tests are intended to measure a single construct. This new paradigm, known as projective IRT, will allow accurate ability estimates for one or more specific dimensions of interest, even when other dimensions are also measured. Such an approach would be quite useful in educational measurement, given the inherently, even if unintentionally, multidimensional nature of most educational assessments.

The proposed three-year project will extend the preliminary work of the research team on projective IRT, extensively test the method through simulation and cross-validation studies, and tackle important practical implementation issues in educational testing. This method will be applied to two large-scale educational assessment data sets, the National Assessment of Educational Progress and the Program for International Student Assessment. The researchers will produce guidance for applied researchers and testing companies, software for conducting projective IRT analysis, and conference presentations and journal articles for disseminating the applied and the technical results of the study.