Bayesian Networks

3

More Complex Cases: Hybrid Bayesian

Networks

In Chapters 1 and 2 we considered BNs with either discrete or continuous

variables. Moreover, in each BN all variables followed probability distributions

belonging to the same family: multinomial or multivariate normal. In the

following, we would like to show that there are no theoretical reasons for such

restrictions and that, according to the phenomenon unde r investigation:

1. we can mix discrete and continuous variables and

2. we can use any kind of distribution.

Unfortunately, this increase in flexibility means BNs become more compli-

cated, and no dedicated R package exists to handle them. Therefore, ...