Construction of Tests

Abstract

The purposes of this chapter are to consider some uses of item response models in the construction of tests and to present some new research results on item selection. With the availability of invariant item statistics, the desirability of item response models for test development work seems clear. But more effective implementation of the models could be achieved if several questions were satisfactorily answered. The choice of a test model is one question that was considered in chapter 8 and will be considered again in the last chapter. It would greatly facilitate the test development process if practical guidelines existed to provide a basis for making this choice. A second question concerns the reasons for item misfit. Several techniques for identifying misfitting items were considered in chapters 8 and 9. At the present level of our technical sophistication, the test developer, faced with a misfitting item, can do little more than subjectively examine the item and hope that the reason for misfit will be apparent.