Methods for Identifying Biased Test Items

This book makes clear to researchers what item-bias methods can (and cannot) do, how they work and how they should be interpreted.

Advice is provided on the most useful methods for particular test situations. The authors explain the logic of each method - from item-response theory to nonparametric, categorical methods - in terms of how differential item functioning (DIF) is defined by the method and how well the method can be expected to work. A summary of findings on the behaviour of indices in empirical studies is included. The book concludes with a set of principles for deciding when DIF should be interpreted as evidence of bias.

This book is highly recommended for all involved in test development or who have any interest in the use of tests in education or in other fields. The necessary mathematics are presented clearly but never obscure the important messages in the book. It will certainly be referred to constantly in my future work in this area. --Educational Research "The fundamental goal of the Measurement Methods for the Social Sciences series is to make complex measurement concepts, topics, and methods available to readers with limited mathematical background but a strong desire to understand, as well as use, methods that are on the forefront of social science assessment. With this book on item bias detection methods, Gregory Camilli and Lorrie Shepard have achieved this goal admirably." --from the Foreword by Richard M. Jaeger What can item bias methods do--and not do--when applied to real test data? Aimed at helping researchers understand how item bias methods work, this book provides practical advice and specific details on the most useful methods for particular testing situations. Beginning with a review of early bias methods and the fairness issues associated with the topic of test bias, the authors explain the logic of each method in terms of how differential item functioning (DIF) is defined by the method--and how well the method can be expected to work in various situations. In addition, chapters include a summary of findings regarding the behavior of the various indexes in empirical studies, especially their reliability, correlation with known bias criteria, and correlations with other bias methods. The book concludes with a set of principles for deciding when DIF should be interpreted as evidence of bias.