Abstract
This project evaluated the adequacy of the time limit on a licensure examination by using residual analysis to detect changes in person speed consistent with a speeded response time profile. Response time data were fit to a Lognormal Response Time Model. The model assumed constant speed, therefore speed changes in a person’s response time profile were expressed in the residuals. The residual analysis identified multiple types of speed changes in a person’s response time profile by using interrupted regression modeling. Regression model coefficients and normative thresholds were used to determine if speed change patterns at regression model change points were consistent with a change from solution behavior to hurried test taking or rapid guessing, and therefore a speeded test profile. The method successfully identified several expected variants of a speeded test profile. The residual analysis increased the number of test takers found to have speeded test profiles by 35% over just considering incomplete tests. With that joint-approach, just 8% of test attempts were speeded, indicating the test time limit was appropriate. Some test attempts had residual profiles which violated analysis assumptions which may have impacted their classification. Despite limitations, the joint-method offered improvement over only considering test incompleteness, and regression methods detected variations of speed increases that other time-based methods for detecting speeded test attempts may not. Detecting speeded test attempts can also help differentiate between suspicious and non-suspicious attempts in test security analysis, and the regression method could be adapted to detect other speed changes such as item preknowledge.