ALIGNING SCALES OF CERTIFICATION TESTS

Abstract

Scores are the most visible and widely used products of a testing program. The choice of score scale has implications for test specifications, equating, and test reliability and validity, as well as for test interpretation. At the same time, the score scale should be viewed as infrastructure likely to require repair at some point. In this report we examine the issue of scale fit—how well the scale fits the intended uses of its scores—for certification tests. Two examples of scale fit are considered: one in which the test has a single threshold that separates the candidate population into pass-fail groups, and one in which the test is required to support a restricted range of multiple thresholds.