Decision Tables and logic in Decision Analysis

Abstract

Unmanageably bushy decision trees result when a decision analysis involves several in vestigations. They can be simplified for riskless tests by deriving the maximum expected utility decision table for the problem as an intermediate step. This table can be logically summarized as Boolean expressions involving the tests. A minimum-cost testing sequence may then be found by manipulation of the Boolean formulas. The relationship between the resulting decision criteria and the receiver operating characteristic is shown.