This course covers several methods for analyzing data with missing values. We start with an introduction to the main ideas using several examples. We then cover weighting methods; imputation methods – hot deck, single and multiple imputation; maximum likelihood for incomplete data distinguishing ignorable and non-ignorable missing data mechanisms; selection and pattern missing models for longitudinal studies; and computationally intensive methods including data augmentation and Gibbs sampling methods. The course will draw heavily from the Little and Rubin book Statistical Analysis with Missing Data. The course is designed for those with at least a master’s level background in statistics. Knowledge of complete data methods including likelihood based methods and familiarity with linear and logistic regression models and repeated measure analyses will be assumed. The major emphasis will be on the basic conceptual issues in dealing with missing data, available methods for analyzing such data, and their applications.
Dr. Little is Richard D. Remington Collegiate Professor of Biostatistics, School of Public Health, & Research Professor, Survey Research Center, Institute for Social Research at the University of Michigan. He is also currently Associate Director for Research & Methodology and Chief Scientist at the United States Census Bureau. His research expertise focuses on handling missing data in a variety of statistical analyses, and inference from sample surveys.