ABSTRACT
Computerized adaptive testing (CAT) broadly refers to tests administered on a computer, where the items are selected so that the overall test-form difficulty is targeted to the proficiency of each examinee. This proficiency targeting results in a more efficient test where greater score or decision accuracy can be achieved using fewer items than in a typical fixed test form (i.e., a test form where every examinee sees exactly the same items). In this context, accuracy is referring to the magnitude of the estimation error variance. A larger error variance implies less stable score estimates. As noted further on in this chapter, another component of accuracy is bias—that is, failure of the estimate to converge to the true (but unknown) parameter of interest.
