SIMULATING THE USE OF DISCLOSED ITEMS IN COMPUTERIZED ADAPTIVE TESTING1,2

Abstract

Regular use of questions previously made available to the public (disclosed) may provide one way to meet the requirement for large numbers of questions in a continuous testing environment, that is, an environment in which testing is offered at test-taker convenience throughout the year rather than on a few prespecified test dates. First it must be shown that such use has effects on test scores small enough to be acceptable. In this study simulations are used to explore the use of disclosed items under a worst case scenario which assumes that disclosed items are always answered correctly. Some item pool and test designs were identified in which the use of disclosed items produces effects on test scores that may be viewed as negligible.