Graphical models: overview

摘要: Graphical Markov models provide a method of representing possibly complicated multivariate dependencies in such way that the general qualitative features can be understood, statistical independencies are highlighted, and some properties derived directly. Variables represented by nodes graph. Pairs may joined an edge. Edges directed if one variable is response to other considered as explanatory, but undirected variables on equal footing. Absence edge typically implies independence, conditional, or marginal depending kind The need for number types graph arises because it helpful represent different kinds dependence structures. Of special importance chain graphs which arranged sequence blocks, any block being footing, joint responses past jointly explanatory future considered. Some main systems outlined, recent research results sketched. Suggestions further reading given. As illustrative example, analysis data treatment chronic pain presented.