ABSTRACT
This chapter presents the new method for detecting test collusion. It aims to fill a void in the literature by presenting a new approach to detect clusters of examinees engaged in test collusion. The B. A. Jacob and S. D. Levitt (JL) method specifically focused on teacher cheating but could conceivably extend to other groups of examinees for which collusion was possible. The JL method identifies collusion through consideration of two summary scores, one relating to the overall atypicality of answer strings within a classroom and the other relating to unexpected score fluctuations. For purposes of detecting collusion, answer similarity indexes are more useful. A simulation study was conducted for purposes of studying the Type I error rate and power, as well as the ability of the collusion detection model to recover known group memberships, under realistic circumstances. Simulating collusion for a group of examinees, however, must proceed a little differently.
