About Bayesian SEM
Bayesian SEM refers to the estimation and evaluation of structural equation models using Bayesian methods. I will not attempt to give a presentation on Bayesian statistics, as there are many such presentations, including those on the main Bayesian tab. The material given here will be aimed at informing those familiar with Bayesian statistics about SEM, as well as providing tutorials on Bayesian SEM for those who wish to implement them.
A nice and very new treatment of winBUGS is -->
Ntzoufras, I 2009. Bayesian Modeling Using WinBUGS. John Wiley & Sons.
It is worth considering what Ntzoufras says on page 84 of his book about using Markov chain Monte Carlo (MCMC) methods such as winBUGS (keep in mind that the BUGS part is Bayesian inference Using Gibbs Sampling).
". . . all WinBUGS users may have noticed the 'health' warnings that always appear in the first page of BUGS and WinBUGS manuals: Beware - Gibbs sampling can be dangerous!" This phrase resembles and reminds us of the health warnings that appear in all cigarette packets. Experienced MCMC users will consent with this comment for two reasons: (1) runnings a Gibbs sampler (and generally an MCMC algorithm) is usually a laborious procedure that requires careful computation of the conditional posterior densities and coding of the correspondng algorithm usng a programming language, and (2) WinGUGS users must be familiar with the basic notions of Bayesian statistical theory and computition; otherwise they may not specify the model correctly, or may interpret the posterior results incorrectly, or might not obtain results from a converged chain or even stick with the first computation problem of the algorithm (e.g., with a simple overflow problem)."
Now, everything you need to get started with WinBUGS can be found at