ISBIS - Dynamic Item Response Modeling and Its Application in Health Economics

Item Response models have been extensively used in Psychometrics, especially in educational programs like ETS (Educational Testing Services) for estimating the learning ability of students. Traditional models include 1PL, 2PL and 3PL models, and their variations for specific applications. In this paper, we consider the problem of estimating the latent trait (i.e. general health condition) based on longitudinal binary measurements using Bayesian hierarchical modeling. We develop three alternative dynamic models where we use Lasso-type shrinkage priors for variable selection, and estimate the underlying model parameters through MCMC iterations. Our proposed models are applied for estimating the overall (latent) health condition of the older individuals. Posterior predictive checks are performed for assessing the reliability of the estimates. Our proposed models will be very useful in different disciplines including health sciences, physical sciences and economics.