Learning structures from data and experts Original Research Article

In modelling complex stochastic systems graphical association models provide a convenient framework. With graphical models the overall structure of association among variables is described in terms of conditional independence, and this structure can be represented graphically. Statistical methods for revealing these basic structures of association on the basis of data and expert knowledge are described. Suggestions on how to make a more detailed modelling will be made, and it will be illustrated how to implement such models in a causal probabilistic network. As an illustration a model for the incidence of fungi attacks and yield in relation to various cultural factors in winter wheat is established.