Winbugs seem to support either stochastic or deterministic relationship between variables. However, many Bayesian Networks represent relationships between variables using conditional probability tables. The "visit to Asia", "burglar alarm", "smoking & cancer" examples are classic introductory material.
However, conditional probability tables used in this approach does not correspond to stochastic or deterministic relationships in Winbugs. Google searches bring examples of Gibbs sampling on Bayesian Networks, but these are mostly algorithms that I'd have to implement in either R or another programming language (for example see: http://www-users.cselabs.umn.edu/classes/Spring-2010/csci5512/notes/gibbs.pdf)
Is there a way of using Winbugs for complex Bayesian Network inference? I need to express causal relationships between different variables (a continuous variable and a discrete one, a categorical one etc), so I need to perform MCMC based inference on hybrid Bayesian networks. I am not so sure if I can use Winbugs for this
Finally, would it make sense to try to transform the conditional probability table based, or causal relationships into a stochastic relationships? If I turn CPT based relationships between variables into regression for example, would that be an acceptable way of performing inference on Bayesian Networks using Winbugs?