ABSTRACT
This chapter introduces readers to Item response theory (IRT) models and methods now being used in personnel assessment and selection. Selecting the item that provides the most information is optimal in a psychometric sense but leads to items with the largest a-parameters being overused across testing sessions, particularly in the early stages of exams when examinees have very similar scores. As more flexible IRT methods are developed, IRT is likely to become the methodology of choice for supporting structured assessments. With modern computer-based testing, response time has become particularly easy to track, and, in noncognitive testing, the number of items answered in less than two seconds can serve an effective flag for careless responding. In the three-parameter logistic model, each item is characterized by an item difficulty parameter, an item discrimination parameter, and a lower asymptote or "guessing" parameter. Historically, the most widely used software for estimating Samejima's Graded Response Model parameters was the multilog computer program.
