The Goal of Equity within and between Computerized Adaptive Tests and Paper and Pencil Forms.

Score comparability is important to those who take tests and those who use them. One important concept related to test score comparability is that of "equity," which is defined as existing when examinees are indifferent as to which of two alternate forms of a test they would prefer to take. By their nature, computerized adaptive tests (CAT) add extra complexity for score comparability issues. At a minimum, comparability between a CAT and a paper-and-pencil (P&P) form means that the score each examinee gets on the CAT and P&P forms should be as interchangeable as possible in terms of the construct being measured and the measurement precision of that construct. The importance of the content issue, termed "content balancing," has been expressed by several researchers. The relative importance or appropriateness of content areas to different regions of the ability scale should guide development of both CAT and P&P forms. Among the approaches suggested for content balancing is a conditional balancing approach advocated by T. C. Davey and L. Thomassen (1995). The simulation study described in this paper extends Davey's approach by including realistic item exposure controls. An artificial 25-item P&P math test was used as a reference and target test for the CAT conditions. Results from this preliminary study seem to confirm the assertion that it is possible to employ item content controls appropriately during item administration of a CAT, following targets based on natural distribution of item content as a function of ability so that total measurement precision is not severely negatively impacted. (Contains 21 figures and 20 references.) (SLD)