Item response theory (IRT) has become the mainstream measurement approach in social sciences concomitant with rising concerns about classical measurement techniques approach estimated item and person parameters. Related to this concern is how comparisons between different test forms could lead to somewhat inconsistent results. With IRT, comparable person estimates can be achieved irrespective of the specific set of items taken. Consequently, results from studies using different test forms can be combined. Also, comparable item estimates can be obtained irrespective of the specific group of subjects taking them. Further, IRT methods place persons and items on the same measurement scale and make comparisons between persons and items feasible. Applications of IRT models are increasingly used in developing a variety of test items in fields such as economics, education, psychology, and medicine. This entry focuses on three binary models: one-parameter logistic/Rasch, two-parameter logistic, and three-parameter logistic models followed by a brief introduction of some widely used polytomous models. Applications of IRT methods in computerized adaptive testing, linking, differential item functioning, item generation, and item banking are then discussed. Finally, a description of some popular software assisting in IRT modeling is provided with examples on how analyses are conducted.
By: Heather A. Handy, Eunbee Kim, Maryam Pezeshki, Yan Yan & Susan Embretson | Edited by: Paul Atkinson, Sara Delamont, Alexandru Cernat, Joseph W. Sakshaug & Richard A. Williams Published: 2020 | Length: 10 | DOI: http://dx.doi.org/10.4135/9781526421036883635 |