Statistical Test Theory for the Behavioral Sciences
Since the development of the first intelligence test in the early 20th century, educational and psychological tests have become important measurement techniques to quantify human behavior. Focusing on this ubiquitous yet fruitful area of research, Statistical Test Theory for the Behavioral Sciences provides both a broad overview and a critical survey of assorted testing theories and models used in psychology, education, and other behavioral science fields. Following a logical progression from basic concepts to more advanced topics, the book first explains classical test theory, covering true score, measurement error, and reliability. It then presents generalizability theory, which provides a framework to deal with various aspects of test scores. In addition, the authors discuss the concept of validity in testing, offering a strategy for evidence-based validity. In the two chapters devoted to item response theory (IRT), the book explores item response models, such as the Rasch model, and applications, including computerized adaptive testing (CAT). The last chapter looks at some methods used to equate tests. Equipped with the essential material found in this book, advanced undergraduate and graduate students in the behavioral sciences as well as researchers involved in measurement and testing will gain valuable insight into the research methodologies and statistical data analyses of behavioral testing.
- Electronic book text | 280 pages
- 31 Aug 2007
- Taylor & Francis Ltd
- Chapman & Hall/CRC
- London, United Kingdom
- 46 Illustrations, black and white
"This book is a comprehensive and well-illustrated overview of the concepts and applications of statistical test theory in the social sciences. ... Various concepts and methodologies are well explained and illustrated through examples that are accompanied with data-based examples. The language of the book is simple. The prerequisite knowledge of mathematics and statistics is kept to a minimum ... The book will undoubtedly appeal to the students and application-oriented researchers in the social sciences who wish to obtain a detailed overview of statistical test theory. This book can be adopted as a textbook in advanced undergraduate and graduate courses in the social sciences and can also be comprehensively used as a part of any course in applied statistics." -Journal of the Royal Statistical Society
Table of contents
PREFACE Measurement and Scaling Definition of a test Measurement and scaling Classical Test Theory True score and measurement error The population of persons Classical Test Theory and Reliability The definition of reliability and the standard error of measurement The definition of parallel tests Reliability and test length Reliability and group homogeneity Estimating the true score Correction for attenuation Estimating Reliability Reliability estimation from a single administration of a test Reliability estimation with parallel tests Reliability estimation with the test-retest method Reliability and factor analysis Score profiles and estimation of true scores Reliability and conditional errors of measurement Generalizability Theory Basic concepts of G theory One-facet designs, the p x i design, and the i : p design The two-facet crossed p x i x j design An example of a two-facet crossed p x i x j design: The generalizability of job performance measurements The two-facet nested p x (i : j) design Other two-facet designs Fixed facets Kinds of measurement errors Conditional error variance Concluding remarks Models for Dichotomous Items The binomial model The generalized binomial model The generalized binomial model and item response models Item analysis and item selection Validity and Validation of Tests Validity and its sources of evidence Selection effects in validation studies Validity and classification Selection and classification with more than one predictor Convergent and discriminant validation: A strategy for evidence-based validity Validation and IRT Research validity: Validity in empirical behavioral research Principal Component Analysis, Factor Analysis, and Structural Equation Modeling: A Very Brief Introduction Principal component analysis (PCA) Exploratory factor analysis Confirmatory factor analysis and structural equation modeling Item Response Models Basic concepts The multivariate normal distribution and polytomous items Item-test regression and item response models Estimation of item parameters Joint maximum likelihood estimation for item and person parameters Joint maximum likelihood estimation and the Rasch model Marginal maximum likelihood estimation Markov chain Monte Carlo Conditional maximum likelihood estimation in the Rasch model More on the estimation of item parameters Maximum likelihood estimation of person parameters Bayesian estimation of person parameters Test and item information Model-data fit Appendix: Maximum likelihood estimation of Î¸ in the Rasch model Applications of Item Response Theory Item analysis and test construction Test construction and test development Item bias or DIF Deviant answer patterns Computerized adaptive testing (CAT) IRT and the measurement of change Concluding remarks Test Equating Some basic data collection designs for equating studies The equipercentile method Linear equating Linear equating with an anchor test A synthesis of observed score equating approaches: The Kernel method IRT models for equating Concluding remarks Answers References Index Each chapter contains an Introduction and Exercises.