Bayesian Networks
By succinctly and effectively translating causal assertions between variables into patterns of probabilistic dependence, Bayesian networks (BNs) facilitate logical and holistic reasoning under uncertainty in complex systems. Such reasoning is necessary for accurate analysis, synthesis, prediction, diagnosis, and decision making. A definition of BNs is first provided and a simple ecological example is introduced that will be used throughout the remainder of the article to illustrate basic concepts. Methods for constructing BNs are next described, including specification of model structure and conditional probabilities, both deliberately and automatically from case data. This is followed by a description of how BNs can be used for prediction, inference, explanation, intervention, and decision. Finally, some special cases of BNs are presented including hierarchical, dynamic, and integrated models. These show how BNs can be used to integrate across scales, disciplines, and levels of complexity.
