 Data augmentation

Data augmentation enlarges a posterior distribution with latent variables whose removal by marginalization recovers the original posterior distribution. Convenient full conditional distributions then permit a Gibbs sampler on the enlarged space.