Multiple imputation under Bayesianly smoothed pattern-mixture models for non-ignorable drop-out
Conventional pattern-mixture models can be highly sensitive to model misspecification. In many longitudinal studies, where the nature of the drop-out and the form of the populat...