Collusion can be indicated when two or more examinees perform more similarly than one would expect after accounting for their ability and the difficulty of the items they answered. In this presentation we introduce a new metric that holds promise for identifying this type of cheating. The metric is based on Yen’s Q3 statistic, can be calculated for tests comprised of dichotomous and/or polytomous items, and can be estimated by widely-available IRT software (e.g., WINSTEPS). Attendees will be shown how the statistic is calculated and provided with explanations of the method’s strengths and weaknesses. Results from preliminary applications of the method were shown in the presentation.
Brett Foley, Ph.D., Alpine Testing Solutions