The calculation of posterior distributions by data augmentation. (English) Zbl 0619.62029

Tanner, Martin A.; Wong, Wing Hung

J. Am. Stat. Assoc. 82, 528-541 (1987).

Data augmentation is a scheme of augmenting observed data to make them easier to analyse; this arises in estimation problems where there are missing values. An algorithm for the Bayesian analysis of augmented data using simulation techniques is presented where the posterior distribution of the parameters of interest is obtained by using the predictive distribution of the augmenting or latent data given the observed data.
The paper is followed by a discussion with contributions from A. P. Dempster, C. N. Morris, D. B. Rubin, S. J. Haberman and A. O’Hagan and a rejoinder by the authors.

Reviewer: P.W.Jones

MSC:

62F15	Bayesian inference
65C05	Monte Carlo methods

Keywords:

Data augmentation; missing values; algorithm; simulation techniques; posterior distribution; predictive distribution; latent data