Item Response Theory

From MaRDI portal





In the years since the publication of the first edition of this book in 1992, see the review Zbl 0761.62161, item response theory (IRT) has continued to be a dynamic field of inquiry. The literature has continued to grow as the breadth of this field has expanded. Offering more than 25\,\(\%\) new material, the second edition clearly describes the most recently utilized IRT models and furnishes detailed explanations of algorithms that can be used to estimate the item or ability parameters under various IRT models -- surrounding issues related to statistical theory, numerical methods, and the mechanisms of computer programs for parameter estimation. Revised and expanded throughout, the second edition provides three new chapters, Chapters 10\,--\,12. A common situation is one where the data set contains multiple, identifiably different groups of examinees who have taken the same instrument. The parameter estimation procedures for this situation allow one to compare groups, differentiated on the basis of some criterion, by a common metric. They also provide means of determining whether items in an instrument exhibit differential item functioning across several groups of examinees. Finally, these procedures can be used to detect item parameter drifts of an instrument. The marginal maximum likelihood estimation procedures for the multiple group situation are presented in Ch. 10, Parameter estimation for multiple group data. A situation of great interest to instrument developers is that of estimating item parameters of instruments which contain items based upon more than one item response model; for example, where dichotomously-scored and polytomously-scored items are intermixed within a given instrument. Developers are interested in this as it adds variety to an instrument and allows them to tailor the item to the characteristics of the topic. Marginal maximum likelihood estimation procedures for a given mixed item type case are presented in Ch. 11, Estimation of item parameters of mixed models. The Gibbs sampler is a member of a new class of procedures known as Markov chain Monte Carlo techniques. Ch. 12, Parameter estimation via Gibbs sampler, presents this relatively new estimation method in IRT. In the first edition, each chapter of the book was accompanied by an appendix containing a BASIC program that implemented the particular procedure. Since the publication of the first edition, the computing power of personal computers has increased dramatically, so that BASIC programs for the three additional chapters are contained in their corresponding appendices. The companion CD contains the BASIC code as well as the Java code for these appendices. The book contains reader-friendly discussions and instructionally oriented material for clear understanding of the computational demands and challenges of IRT estimation procedures. From the contents: The item characteristic curve; dichotomous response. Estimating the parameters of an item characteristic curve. Maximum likelihood estimation of examinee ability. Procedures for estimating both ability and item parameters. The Rasch model. Parameter estimation via marginal maximum likelihood estimation and an expectation-maximization algorithm. Bayesian parameter estimation procedures. The graded item response. Nominally scored items. Parameter estimation for multiple group data. Estimation of item parameters of mixed models. Parameter estimation via Gibbs sampler.




Cited in
(only showing first 100 items - show all)


Describes a project that uses

Uses Software






This page was built for publication: Item Response Theory

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4821962)