Key points are not available for this paper at this time.
SUMMARY Models with complex structure arise in many social science applications and appear natural candidates for the use of Markov chain Monte Carlo methods for inference. Conditional independence assumptions simplify the model specification and make estimation using Gibbs sampling particularly appropriate. Two examples are discussed: random effects models for repeated ordered categorical data and sensitivity analysis to assumptions concerning the mechanism underlying informative drop-out in a longitudinal study. The use of a program BUGS is demonstrated.
Best et al. (Mon,) studied this question.