Bayesian inference using Gibbs sampling in applications and curricula of decision analysis

摘要

Purpose: To explore and highlight Bayesian inference using Gibbs sampling (BUGS) software, which uses MCMC methods, to numerically obtain posteriors for nonconjugate priors. Summary: By using decision maker's true non-conjugate belief, the study suggests that distribution that leads to optimal decisions. Other methods exist that can use non-conjugate priors, but they are to be implemented ad hoc, in the absence of any supporting software. Bayesian inference using Gibbs sampling is useful in a wide range of decision problems and is easy to implement. Hence it is a valuable decision tool, deserving a place in the curricula. (29 refs.)