Missing Data in Longitudinal Studies

This book provides a unified Bayesian approach to handle missing data in longitudinal studies. It contains examples and case studies on schizophrenia, aging, HIV, and smoking cessation. The authors describe assumptions that include MAR and ignorability, demonstrate the importance of covariance modeling with incomplete data, and cover mixture and selection models for nonignorable missingness. They also present methods for representing untestable assumptions using prior distributions. Several analyses deal with nonignorable missingness as well as illustrate the models and methods. Many analyses are implemented using WinBUGS, with the code provided on a supplementary web page.