METRI and The Allowance List Problem

The fundamentals of the METRI approach to model-building are discussed. This approach may have applicability where the system to be modelled involves very large numbers of variables, ignorance, uncertainty, and hierarchical relationships, and where the decisions to be made involve enormous amounts of detail. Since the approach was developed in the context of the “allowance list” problem, the application of the approach to that problem is discussed in detail. The problem of “validation” is also discussed, since, as in some other cases, the payoff measure defined by the model has no close parallel in terms of independently available data. The paper is basically an account of certain aspects of a U.S. Navy project.