Applied Missing Data Analysis in the Health Sciences - Xiao-Hua Zhou

Résumé :

A modern and practical guide to the essential concepts and ideas for analyzing data with missing observations in the field of biostatistics

With an emphasis on hands-on applications, Applied Missing Data Analysis in the Health Sciences outlines the various statistical methods for the analysis of missing data. The authors acknowledge the limitations of established techniques and provide newly-developed methods with concrete applications in areas such as causal inference.

Organized by types of data, chapter coverage begins with an overall introduction to the existence and limitations of missing data and continues into techniques for missing data inference, including likelihood-based, weighted GEE, multiple imputation, and Bayesian methods. The book subsequently covers cross-sectional, longitudinal, hierarchical, survival data. In addition, Applied Missing Data Analysis in the Health Sciences features:

• Multiple data sets that can be replicated using SAS^(R), Stata^(R), R, and WinBUGS software packages
• Numerous examples of case studies to illustrate real-world scenarios and demonstrate applications of discussed methodologies
• Detailed appendices to guide readers through the use of the presented data in various software environments

Applied Missing Data Analysis in the Health Sciences is an excellent textbook for upper-undergraduate and graduate-level biostatistics courses as well as an ideal resource for health science researchers and applied statisticians.

Biographie:
Xiao-Hua Zhou, PhD, is Professor in the Department of Biostatistics at the University of Washington and Director and Research Career Scientist at the Biostatistics Unit of the Veterans Affairs Puget Sound Health Care System. Dr. Zhou is Associate Editor of Statistics in Medicine and has published over 200 journal articles in his areas of research interest, which include statistical methods in diagnostic medicine, analysis of skewed data, causal inferences, and statistical methods for assessing predictive values of biomarkers. Chuan Zhou, PhD, is Research Associate Professor of Biostatistics in the Department of Pediatrics at University of Washington. He has coauthored numerous journal articles in his research areas of interest, which include clinical trials, health service research, diagnostics, missing data, and causal inference. Danping Liu, PhD , is Investigator in the Division of Intramural Population Health Research at the Eunice Kennedy Shriver National Institute of Child Health and Human Development. He has authored numerous research articles in his research areas of interest, which include medical diagnostic testing and ROC curve, missing data methodologies, longitudinal data analysis, and non- and-semi-parametric inferences. Xiaobo Ding, PhD, is Assistant Professor in the Academy of Mathematics and Systems Science at the Chinese Academy of Sciences. His research interests include dimension reduction, variable selection, missing data, confidence bands, and goodness of fit tests.

Sommaire:
List of Figures xv List of Tables xvii Preface xix Introduction xxi 1 Missing Data Concepts and Motivating Examples 1 1.1 Overview of Missing Data Problem 1 1.2 Mechanisms 3 1.3 Data examples 8 2 Overview of Methods for Dealing with Missing Data 19 2.1 Methods that remove observations 20 2.2 Methods that utilize all available data 21 2.3 Methods that impute missing values 22 3 Design Considerations in the Presence of Missing Data 31 3.1 Design factors related to missing data 32 3.2 Strategies for limiting missing data in the design of clinical trials 33 3.3 Strategies for limiting missing data in the conduct of clinical trials 34 3.4 Minimize the impact of missing data 35 3.5 Sample size and power consideration in the presence of missing data 36 4 Cross-sectional Data Methods 41 4.1 Overview of General Methods 41 4.2 Data Examples 42 4.3 Maximum Likelihood Approach 44 4.4 Bayesian Methods 61 4.5 Multiple Imputation 71 4.6 Inverse Probability Weighting 76 4.7 Weighted Estimating Equation Approaches 79 4.8 Doubly Robust Estimators 80 4.9 Additional Theories 83 5 Longitudinal Data Methods 97 5.1 Overview of Chapter 97 5.2 Examples 98 5.3 Longitudinal Regression Models for Complete Data 101 5.4 Missing Data Settings and Simple Methods 111 5.5 Likelihood Approach 112 5.6 Weighted GEE (WEE) with MAR Dropout 117 5.7 Extension to Nonmonotone Missingness 123 5.8 Multiple Imputation (MI) 125 5.9 Bayesian Inference 139 5.10 Other Approaches 141 5.11 Appendix: Technical Details 149 6 Survival Analysis under Ignorable Missingness 153 6.1 Overview of the chapter 153 6.2 Introductions 154 6.3 Enhanced complete-case analysis 157 6.4 Weighted methods 159 6.5 Imputation methods 168 6.6 Nonparametric maximum likelihood estimation 171 6.7 Transformation model 172 6.8 Pathways study 174 6.9 Concluding remarks 175 7 Nonignorable Missingness 177 7.1 Introduction 177 7.2 Cross-sectional data: selection model 179 7.3 Longitudinal data with dropout 180 7.4 Bayesian analysis for GLMs 191 7.5 Multiple imputation 195 7.6 Inverse probability weighted methods 199 8 Analysis of Randomized Clinical Trials with Non-Compliance 215 8.1 Overview of the chapter 215 8.2 Examples 217 8.3 Some Common but Naive Methods 218 8.4 Notations, Assumptions, and Causal Definitions 220 8.5 Method of Instrumental Variables 223 8.6 Another Moment-based Method 224 8.7 Maximum Likelihood and Bayesian Method 227 8.8 Noncompliance and Missing Some Outcome Data 232 8.9 Analysis of the Two Examples 241 8.10 Other Methods for Dealing with both Noncompliance and Missingdata 242 8.11 Appendix: Multivariate Delta Method 243