Measurement Models for Psychological Attributes

Despite the overwhelming use of tests and questionnaires, the psychometric models for constructing these instruments are often poorly understood, leading to suboptimal measurement. Measurement Models for Psychological Attributes is a comprehensive and accessible treatment of the common and the less than common measurement models for the social, behavioral, and health sciences. The monograph explains the adequate use of measurement models for test construction, points out their merits and drawbacks, and critically discusses topics that have raised and continue to raise controversy. Because introductory texts on statistics and psychometrics are sufficient to understand its content, the monograph may be used in advanced courses on applied psychometrics and is attractive to both researchers and graduate students in psychology, education, sociology, political science, medicine and marketing, policy research, and opinion research. The monograph provides an in-depth discussion of classical test theory and factor models in Chapter 2; nonparametric and parametric item response theory in Chapter 3 and Chapter 4, respectively; latent class models and cognitive diagnosis models in Chapter 5; and discusses pairwise comparison models, proximity models, response time models, and network psychometrics in Chapter 6. The chapters start with the theory and methods of the measurement model and conclude with a real-data example illustrating the measurement model.

Table of Contents Acknowledgments Glossary of Notion and Acronyms 1. Measurement in the Social, Behavioral, and Health Sciences Introduction Methodological Procedures and Psychometric Measurement Models Relation of Measurement Model Attribution Scale Developing Attribute Theory is Important Measurement Instruments Measurement Models Scales of Measurement Causes of Messy Data A Scale for Transitive Reasoning Cycle of Instrument Construction This Monograph 2. Classical Test Theory and Factor Analysis Historical Introduction The Classical Test Method Measurement Level and Norm Scores Model Assumptions Repeatability of Test Scores: Reliability Methods for Estimating Reliability Methods Commonly Used in Test-Construction Practice Parallel-test method Retest Method Split-Half Method Internal Consistency Method Reliability Methods Based on One Test Administration Method Method Method Method Method Method Greatest Lower Bound Special Topics Concerning Methods through and the GLB Mutual Relationships of Lower Bounds and Reliability Discrepancy of Methods through and the GLB Overestimation of Reliability in Real Data Confidence intervals Reliability versus Measurement Precision Traditional Methods Alternative Methods and Special Topics Constructing Scales in the Classical Test Theory Context Corrected Item-Total Correlations and Oblique Multiple Group Method Principal Component Analysis Factor Analysis Factor-analysis approach to reliability One-Factor Model Multi-Factor Model Real-Data Example: The Type D Scale14 (DS14) Discussion 3. Nonparametric Item Response Theory and Mokken Scale Analysis Introduction Model of Monotone Homogeneity Prerequisites Definitions and Notation Assumptions Strict and Essential Unidimensional IRT An Ordinal Scale for Person Measurement Goodness of Fit Methods Unidimensionality: Scalability and Item Selection Scalability Coefficients and Scale Definition Modified Scalability Bounds Mokken’s Automated Item Selection Procedure Modified Procedure to Produce Maximum-Length Scales Sample Size and Concluding Remarks Monotonicity Binning Order-Restricted Likelihood Ratio Test Kernel Smoothing Polytomous-Item Monotonicity Local Independence The CA Method The DETECT Method Comparative Research Data Example: The Type D Scale14 (DS14) Revisited Using Nonparametric IRT Model of Double Monotonicity Goodness of Fit Methods Method Manifest Invariant Item Ordering Other Methods for Investigating an Invariant Item Ordering Reliability Data Example: The Type D Scale14 (DS14) Continued Discussion 4. Parametric Item Response Theory and Structural Extensions Introduction A Taxonomy for IRT Models Some Basic IRT Models for Dichotomous Items Guttman Model Normal-Ogive Models 1-Parameter Logistic Model or Rasch Model The Model, Separability of Parameters Sufficiency and Estimation Information Functions and Measurement Precision Goodness of Fit Methods The Rasch Paradox Epilogue 2 and 3-Parameter Logistic Models Some Basic IRT Models for Polytomous Items Adjacent Category Models Cumulative Probability Models Continuation Ratio Models Filling in the Taxonomy IRT Models for Special Purposes Linear Logistic Model Generalized Rasch Model with Manifest Predictors Multidimensional IRT Models Data Example: Transitive Reasoning Discussion 5. Latent Class Models and Cognitive Diagnostic Models Introduction Latent Class Model An Example: Proportional Reasoning by means of the Balance Scale Introduction The Unrestricted Model Restricted Models Estimation Goodness of Fit Methods Likelihood Statistic Assessing Individual Items Information Fit Measures Special Topics Ordered LCM and Testing Monotonicity in Nonparametric IRT Data Example: Proportional Reasoning by means of the Balance Scale Discussion Cognitive Diagnostic Model An example: Identifying Patients’ Disorder Profiles Using the MCMI-III Introduction Models Deterministic Input, Noisy "AND" Gate Model Reduced Reparametrized Unified Model Deterministic Input, Noisy "OR" Gate Model General Diagnostic Model Generalized-DINA or G-DINA Model Log-Linear Cognitive Diagnostic Model Estimation Goodness of Fit Methods Absolute Fit Assessment Relative Fit Assessment Relationship to Nonparametric IRT Data Example: Identifying Patients’ Disorder Profiles Using the MCMI-III Discussion General Discussion 6. Pairwise Comparison, Proximity, Response Time, and Network Models Introduction Pairwise Comparison Models Thurstone Model Bradley-Terry-Luce Model Discussion Proximity Models Deterministic Model Probabilistic Models Discussion Response Time Models Lognormal Model Diffusion Model Discussion Network Psychometrics Network Approach for Gaussian Data Prerequisites for Gaussian Data Networks Networks for Gaussian Data Network Approach for Binary Data Discussion References

Klaas Sijtsma is a professor of Methods of Psychological Research at the Tilburg School of Social and Behavioral Sciences, Tilburg University, the Netherlands. His research specializes in psychometrics, in particular, all issues related to the measurement of psychological attributes by means of tests and questionnaires. He is a past President of the Psychometric Society, editorial board member for several journals, and has authored two other books on measurement. Andries L. van der Ark is professor of Psychometrics at the Research Institute of Child Development and Education, Faculty of Social and Behavioural Sciences, University of Amsterdam, the Netherlands. His primary research interests include reliability analysis, nonparametric item response theory, and categorical data analysis. The authors have published over 40 papers together on measurement in the social and behavioral sciences.