Interactive Graphical Aids for Bayesian Hierarchical Inference.

摘要

Bayesian hierarchical inference is an optimum methodology for diagnosing the probable state of an unknown variable from data which are known, but which are known, but which are only indirectly related to the variable of interest. The present report describes a general procedure for Bayesian hierarchical inference implemented on an interactive graphics computer. It allows an intelligence analyst to build a hierarchical inference tree structure, assess the required probability distributions, and compute the likelihood distribution over the target variable. The procedure allows variables to be either, discrete, such as the type of research being pursued by a foreign government, or continuous, such as the personnel strength of a foreign tank regiment. (Author)