Applied Missing Data Analysis, Second Edition (Methodology in the Social Sciences)

Description

The most user-friendly and authoritative resource on missing data has been completely revised to make room for the latest developments that make handling missing data more effective. The second edition includes new methods based on factored regressions, newer model-based imputation strategies, and innovations in Bayesian analysis. State-of-the-art technical literature on missing data is translated into accessible guidelines for applied researchers and graduate students. The second edition takes an even, three-pronged approach to maximum likelihood estimation (MLE), Bayesian estimation as an alternative to MLE, and multiple imputation. Consistently organized chapters explain the rationale and procedural details for each technique and illustrate the analyses with engaging worked-through examples on such topics as young adult smoking, employee turnover, and chronic pain. The companion website includes datasets and analysis examples from the book, up-to-date software information, and other resources.   New to This Edition *Explanded coverage of Bayesian estimation, including a new chapter on incomplete categorical variables. *New chapters on factored regressions, model-based imputation strategies, multilevel missing data handling methods, missing not at random analyses, and other timely topics. *Presents cutting-edge methods developed since the 2010 first edition; includes dozens of new data analysis examples. *Most of the book is entirely new.   Cover Half Title Page Series Editor Title Page Copyright Series Editor’s Note Preface Contents 1. Introduction to Missing Data 1.1 Chapter Overview 1.2 Missing Data Patterns 1.3 Missing Data Mechanisms 1.4 Diagnosing Missing Data Mechanisms 1.5 Auxiliary Variables 1.6 Analysis Example: Preparing for Missing Data Handling 1.7 Older Missing Data Methods 1.8 Comparing Missing Data Methods via Simulation 1.9 Planned Missing Data 1.10 Power Analyses for Planned Missingness Designs 1.11 Summary and Recommended Readings 2. Maximum Likelihood Estimation 2.1 Chapter Overview 2.2 Probability Distributions versus Likelihood Functions 2.3 The Univariate Normal Distribution 2.4 Estimating Unknown Parameters 2.5 Getting an Analytic Solution 2.6 Estimating Standard Errors 2.7 Information Matrix and Parameter Covariance Matrix 2.8 Alternative Approaches to Estimating Standard Errors 2.9 Iterative Optimization Algorithms 2.10 Linear Regression 2.11 Significance Tests 2.12 Multivariate Normal Data 2.13 Categorical Outcomes: Logistic and Probit Regression 2.14 Summary and Recommended Readings 3. Maximum Likelihood Estimation with Missing Data 3.1 Chapter Overview 3.2 The Multivariate Normal Distribution Revisited 3.3 How Do Incomplete Data Records Help? 3.4 Standard Errors with Incomplete Data 3.5 The Expectation Maximization Algorithm 3.6 Linear Regression 3.7 Significance Testing 3.8 Interaction Effects 3.9 Curvilinear Effects 3.10 Auxiliary Variables 3.11 Categorical Outcomes 3.12 Summary and Recommended Readings 4. Bayesian Estimation 4.1 Chapter Overview 4.2 What Makes Bayesian Statistics Different? 4.3 Conceptual Overview of Bayesian Estimation 4.4 Bayes’ Theorem 4.5 The Univariate Normal Distribution 4.6 MCMC Estimation with the Gibbs Sampler 4.7 Estimating the Mean and Variance with MCMC 4.8 Linear Regression 4.9 Assessing Convergence of the Gibbs Sampler 4.10 Multivariate Normal Data 4.11 Summary and Recommended Readings 5. Bayesian Estimation with Missing Data 5.1 Chapter Overview 5.2 Imputing an Incomplete Outcome Variable 5.3 Linear Regression 5.4 Interaction Effects 5.5 Inspecting Imputations 5.6 The Metropolis–Hastings Algorithm 5.7 Curvilinear Effects 5.8 Auxiliary Variables 5.9 Multivariate Normal Data 5.10 Summary and Recommended Readings 6. Bayesian Estimation for Categorical Variables 6.1 Chapter Overview 6.2 Latent Response Formulation for Categorical Variables 6.3 Regression with a Binary Outcome 6.4 Regression with an Ordinal Outcome 6.5 Binary and Ordinal Predictor Variables 6.6 Latent Response Formulation for Nominal Variables 6.7 Regression with a Nominal Outcome 6.8 Nominal Predictor Variables 6.9 Logistic Regression 6.10 Summary and Recommended Readings 7. Multiple Imputation 7.1 Chapter Overview 7.2 Agnostic versus Model-Based Multiple Imputation 7.3 Joint Model Imputation 7.4 Fully Conditional Specification 7.5 Analyzing Multiply Imputed Data Sets 7.6 Pooling Parameter Estimates 7.7 Pooling Standard Errors 7.8 Test Statistic and Confidence Intervals 7.9 When Might Multiple Imputation Give Different Answers 7.10 Interaction and Curvilinear Effects Revisited 7.11 Model-Based Imputation 7.12 Multivariate Significance Tests 7.13 Summary and Recommended Readings 8. Multilevel Missing Data 8.1 Chapter Overview 8.2 Random Intercept Regression Models 8.3 Random Coefficient Models 8.4 Multilevel Interaction Effects 8.5 Three-Level Models 8.6 Multiple Imputation 8.7 Joint Model Imputation 8.8 Fully Conditional Specification Imputation 8.9 Maximum Likelihood Estimation 8.10 Summary and Recommended Readings 9. Missing Not at Random Processes 9.1 Chapter Overview 9.2 Missing Not at Random Processes Revisited 9.3 Major Modeling Frameworks 9.4 Selection Models for Multiple Regression 9.5 Model Comparisons and Individual Influence Diagnostics 9.6 Selection Model Analysis Examples 9.7 Pattern Mixture Models for Multiple Regression 9.8 Pattern Mixture Model Analysis Examples 9.9 Longitudinal Data Analyses 9.10 Diggle–Kenward Selection Model 9.11 Shared Parameter (Random Coefficient) Selection Model 9.12 Random Coefficient Pattern Mixture Models 9.13 Longitudinal Data Analysis Examples 9.14 Summary and Recommended Readings 10. Special Topics and Applications 10.1 Chapter Overview 10.2 Descriptive Summaries, Correlations, and Subgroups 10.3 Non-Normal Predictor Variables 10.4 Non-Normal Outcome Variables 10.5 Mediation and Indirect Effects 10.6 Structural Equation Models 10.7 Scale Scores and Missing Questionnaire Items 10.8 Interactions with Scales 10.9 Longitudinal Data Analyses 10.10 Regression with a Count Outcome 10.11 Power Analyses for Growth Models with Missing Data 10.12 Summary and Recommended Readings 11. Wrap-Up 11.1 Chapter Overview 11.2 Choosing a Missing Data-Handling Procedure 11.3 Software Landscape 11.4 Reporting Results from a Missing Data Analysis 11.5 Final Thoughts and Recommended Readings Appendix: Data Set Descriptions References Author Index Subject Index About the Author Notation Guide