Co-Principal Investigator: Xu, Gongjun

When the constructs that educational assessments try to measure are multifaceted, whether by design or not, multidimensional item response theory (MIRT), also known as item factor analysis, provides a unified framework and convenient statistical tool for item analysis, calibration, and scoring. The advancement of computational and statistical techniques has allowed for increased usage of MIRT models, but even with state-of-the-art algorithms, the computation can still be time-consuming, especially when the number of factors is at least five. The purpose of this grant is to develop more efficient statistical models, software, and guidelines for high-dimensional data, large sample sizes, large item banks, and complex designs.

After theoretical work to develop the statistical models, the research team will conduct simulation studies based on real data in order to test the properties of the new models and to compare them to currently used models. The team will use data from NELS 88 and NAEP to provide an applied demonstration of the new models. They will also develop software for estimating the model parameters and for detecting differential item functioning (DIF) within the MIRT framework, with a plan to post the source code on GitHub. The software will have a more user-friendly form via a Shiny app, which will have a user's manual that documents detailed examples and recommended guidelines for students, researchers, and assessment practitioners. The team will use online modules and in-person workshops to test and the user-friendliness of the software and make refinements as needed. The software and the theoretical work will be disseminated through peer-reviewed journal articles and conference presentations.