Title: Multidimensional item response theory applications with mirt and mirtCAT

Author: Phil Chalmers

Affiliation: York University

Abstract: Item response theory (IRT) is a latent variable framework
designed for modeling educational and psychological measurements.  Popular
representations of IRT include the Rasch, partial credit, graded response,
and three-parameter logistic model.  Today, IRT has become the state of the
art in a variety of scientific disciplines ranging from large scale
educational assessments, like the PISA study and standardized psychological
testing, to the development of objective scales for measuring, e.g, the
degree of physical impairment in medical research.

The mirt and mirtCAT packages are currently being developed to help users
analyze testing response data, study IRT models via Monte Carlo
simulations, and implement real-time testing interfaces in R. The mirt
package has been design as a general purpose IRT package for estimating
item- and group-level parameters, diagnosing item and test misfit,
detecting differential item and test functioning, modeling explanatory
covariate terms, and scoring tests using latent variable modeling
approaches.  The focus of mirtCAT, on the other hand, is to provide tools
for multidimensional computerized adaptive testing (MCAT) methodology by
providing functions to generate graphical user interfaces, as well as to
perform Monte Carlo simulations for MCAT designs.  Both packages are based
upon the same underlying framework, and provide a fluid workflow between
collecting and analyzing item response data.

During this presentation, I will briefly demonstrate various aspects of
mirt for analyzing response data with IRT models, and give various live
presentations of how MCATs can be build and summarized with mirtCAT.
Several examples will be presented using real and simulated datasets, and
future work for the packages will be discussed.