|
Abstract:
|
Much of the current Item-Response Theory (IRT) literature assumes that a person's ability to correctly answer a particular question is constant throughout the test. However, if examinees tire during the course of the test, these standard models will produce biased estimates of the item characteristics for items that appear near the end of the exam. We propose a new class of Item-Response Theory (IRT) models that can capture an individual's diminishing capacity to answer test questions, and allow us to both provide more accurate estimates for the item-characteristics, and analyze potential relationships between individual fatigue patterns and the IRT parameters. We then apply these models to data from the National Board of Medical Examiners, corresponding to 8-hour long exams in which the effect of fatigue is significant.
|
