The Best Posterior Probability Model Could Be Different from the Actual Causal Model
Kruger JW 2011 The Causal Structure Could be Different from the Best Posterior Probability Structure If It Is Not a Tree Structure, Proceedings of the 2011 ORSSA Annual Conference: 11-18
6 Pages Posted: 28 Jul 2012
Date Written: September 18, 2011
Abstract
An example is shown that the best posterior probability model, as found by the Cooper and Herskovits equation, is not always the causal model. This involves moving from a causal model, to a frequency distribution, to a correlation matrix and back to a model. The elimination heuristic, utilising independence relations, is introduced as a first step towards getting the actual causal model. At this stage this method is not developed enough to find the causal model.
Keywords: Bayesian Belief Networks, Causal Model, Posterior probability
JEL Classification: C11, C44
Suggested Citation: Suggested Citation