TITLE: Statistical Analysis with Missing Data
INSTRUCTOR: Prof. Rod Little, University of Michigan, and Prof. Qixuan Chen, Columbia University
MODERATOR: Alfred H. Balch
Abstract:
Missing data are a common challenge in health and social science research. Statistical methods and tools can be used to handle missing data to achieve valid statistical inference. This short course will integrate the principle concepts and methods commonly used in statistical analysis with missing data and their applications in surveys, longitudinal studies, and clinical trials. On Day 1 of the course, we first define missing data, patterns of missing data, and missing mechanisms. We then introduce the weighted complete-case analysis, reviewing methods for creating weights and conducting weighted analysis. Finally, we will present incomplete data analysis using maximum likelihood and Bayes methods. On Day 2, we first discuss using multiple imputation (MI) to handle missing data. We cover both implicit and explicit MI procedures. We then discuss likelihood methods for missing not at random (MNAR) models, based on both selection and pattern-mixture models. Finally, we discuss missing data in clinical trials.
The short course will integrate lectures with case studies and hands-on computer lab sessions to put concepts into practice. We include a wide variety of examples to illustrate the techniques and approaches, focusing on weighting in surveys on Day 1 and on MI by chained equations and sensitivity analysis under MNAR on Day 2.
Day 1 Topics Covered
- Introduction and overview
- Complete-case analysis
- Methods for creating weights
- Inverse-probability weighting
- Doubly-robust methods
- Likelihood methods with incomplete data
- Bayes for missing data
Day 2 Topics Covered
- Multiple imputation inference
- Hot-deck imputation
- MI based on joint distribution of incomplete data
- MI using chained equations
- Likelihood methods for MNAR models
- Pattern-mixture models
- Missing data in clinical trials
Instructors’ Biography:
Roderick J. Little is Richard D. Remington Distinguished University Professor of Biostatistics at the University of Michigan, where he also holds appointments in the Department of Statistics and the Institute for Social Research. From 2010-21012 he was the inaugural Associate Director for Research and Methodology and Chief Scientist at the U.S. Census Bureau. He has over 250 publications, notably on methods for the analysis of data with missing values and model-based survey inference, and the application of statistics to diverse scientific areas, including medicine, demography, economics, psychiatry, aging and the environment. His book “Statistical Analysis with Missing Data” with Donald Rubin is now in its 3rd edition, and has over 30,000 google scholar citations. Little is an elected member of the International Statistical Institute, a Fellow of the American Statistical Association and the American Academy of Arts and Sciences, and a member of the Institute of Medicine of the U.S. National Academies. In 2005, Little was awarded the American Statistical Association’s Wilks Medal for research contributions, and he gave the President’s Invited Address at the Joint Statistical Meetings. He was the COPSS Fisher Lecturer at the 2012 Joint Statistics Meetings.
Qixuan Chen is Associate Professor of Biostatistics at Columbia University. Her research focuses on statistical methods development for handling missing data and measurement error arising from health studies. She has also made important contributions in developing novel methods for the analysis of complex survey data. She has been actively engaged in building analysis tools to promote the use of novel statistical methods in health research, with applications to environmental health sciences, psychiatry and mental health, substance abuse, and traffic safety. She is an Associate Editor for Biometrics.