Models and Estimation Procedures for the Analysis of Subjects-by-Items Data Arrays

This final report describes results obtained in a study of models and estimation procedures for the analysis of subjects-by-items data arrays. The results include a new transformation to render an array additive and some simple empirical Bayes methods for estimating subject and item marginal effects. Unfortunately, the problems which these results concern are not as central to the problems of model-based psychological measurement as I originally thought. The report discusses a new problem which I began working on roughly midway through this project, a problem which is crucial for the foundations of model-based mastery testing.