Bayesian Selection of Decomposable Models With Incomplete Data
Paola Sebastiani, Marco F. Ramoni · Journal of the American Statistical Association · 2001
This article describes a new approach to Bayesian selection of decomposable models with incomplete data. This approach requires the characterization of new ignorability conditions for the missing-data mechanism and the development of new computational methods. Both issues are considered, and solutions are proposed. Theory and methods are assessed in controlled experiments and in the analysis of one real-life incomplete dataset.