|
Abstract:
|
Models for analyzing multivariate data sets with missing values require strong, often unassessable, assumptions. The most common of these is that the mechanism that created the missing data is ignorable - a twofold assumption dependent on the inferential procedure used. The first part, under the Bayesian and likelihood paradigms requires that the missing data are missing at random (MAR), an assumption that is impossible to test using the observed data alone. Combining the slightly stronger assumption that the missing data mechanism is missing always at random (MAAR) and the data are row exchangeable however, leads to the weakest possible testable assumption: the missing data indicators only depend on the fully observed outcome variables. In this paper we propose three different diagnostics methods for testing the validity of these assumptions. Although MAAR is not a necessary condition to ensure validity under the Bayesian and likelihood paradigms it is a sufficient condition and should serve as a caution; encouraging the statistician to carry out sensitivity analysis.
|
