ABSTRACT
Over the past 30 years, a proliferation of item response models has emerged. In the Rasch logistic item response model family, notably, the simple logistic model (Rasch, 1980; Volume One, Chapter 3), the partial credit model (Volume One, Chapter 7), the rating scale model (Volume One, Chapter 5), the facets model (Linacre, 1989) and the linear logistic model (Fischer, 1973; Volume One, Chapter 13) have all played an important role in the analysis of item response data. Similarly, the two-parameter logistic model has been extended to polytomous items as the generalized partial credit model (Volume One, Chapter 8) and a variety of multidimensional models have been proposed as extensions for both the Rasch and two-parameter model families (Embretson, 1992; Volume One, Chapter 14; Reckase, 2009; Volume One, Chapter 12). Typically, the development of parameter estimation procedures is specific to each model, and the same some held for the development of dedicated software programs. Surveying the family of Rasch models, Adams and Wilson (1996) developed a unified approach to specifying the models and then consequentially estimating the parameters. There are at least two advantages of developing one single framework to encompass a family of models. First, the development of the estimation procedures and associated software for the implementation of the models can be streamlined within a single framework of models. That is, one only needs to develop one's set of estimation procedures and software program to carry out the estimation of the parameters in the models. Second, a generalized framework provides an opportunity for the development of new models that can fit within the framework. This allows for the flexible application of item response models to suit users' requirements.
