A METHOD FOR COMPUTING PROBABILITIES IN COMPLEX SITUATIONS.

The thesis presents a solution to the problem of describing conveniently and precisely complex probabilistic relationships between events. While the use of a large probability space of joint events is a theoretical solution, it is often useless in practical problems because it is too far removed from the model builders understanding of the problem. Furthermore, in reasonably large problems, such as medical diagnosis, the joint probability space is so large that storing the probabilities of joint events or working with them is impossible. The method developed is a combination of probability and Boolean algebra. Both graphical and algebraic notations are presented. The method allows models to be constructed using the notion of cause and effect. Cause and effect is not the only interpretation a physical analog and a mathematical approach based on probability and Boolean algebra are presente as alternatives. Equally important is the algorithm for computing probabilities. The algorithm uses the structure of the model and theorems of probability to simplify the computation. Author