Abstract
This article describes the computation of and sampling from the posterior density for censored regression problems with normal and generalized log-gamma errors. The data augmentation algorithm (Tanner and Wong 1987) is facilitated in the normal error case because of the form of the augmented posterior. In the generalized log-gamma context, this simplicity is absent. The work of Sweeting (1981) is used as a motivation to develop an importance sampling scheme to sample from an augmented posterior. It is shown how the predictive distribution for a new observation may be computed and sampled from. The methodology is illustrated with two examples.
Original language | English (US) |
---|---|
Pages (from-to) | 829-839 |
Number of pages | 11 |
Journal | Journal of the American Statistical Association |
Volume | 85 |
Issue number | 411 |
DOIs | |
State | Published - Sep 1990 |
Keywords
- Bayesian inference
- Data augmentation
- Multiple imputation
- Simulation
ASJC Scopus subject areas
- Statistics and Probability
- Statistics, Probability and Uncertainty