Ignorable dropout in longitudinal studies BY MARIA

This paper provides a concise de nition for ignorable dropout. This is done primarily from a frequentist perspective. The de nition of ignorable dropout depends on both the population of inference and the type of statistical methodology used for inference. Di erent types of dropouts are described and compared to those found in the literature. Ignorability conditions are then given for the following types of inference: Likelihood-based Inference, Preservation of Marginal Moments and Preservation of Conditional Moments.