chapter  19
24 Pages

Priors on Hypergraphical Models via Simplicial Complexes

It is common to model the joint probability distribution of a family of n random variables {X1, . . . , Xn} in two stages: First to specify the conditional dependence structure of the distribution, then to specify details of the conditional distributions of the variables within that structure [3, 7]. The structure may be summarized in a variety of ways in the form of a graph G = (V , E) whose vertices V = {1, ..., n} index the variables {Xi} and whose edges E ⊆ V × V in some way encode conditional dependence.