ABSTRACT
An item response theory (IRT) model needs to fit the data if the benefits of IRT and the model are to be realized. Selecting an appropriate model given the data is based, at least in part, on model-data fit. In general, a saturated model will fit the data at least as well if not better than a less saturated model. However, a model that is more complex than appropriate violates the principle of parsimony. That is, the simplest model should be selected that still provides a useful explanation of the data. In general, a more parsimonious model will have less chance of introducing inconsistencies, ambiguities, and redundancies into the explanation. The objective in model selection is to select a model that provides the best fit to the data but also has the capability for predicting future or different data.
