DATA UNCERTAINTY, MODEL AMBIGUITY, AND MODEL IDENTIFICATION

This paper is based on a talk presented at the 1986 Resource Modeling Conference in Newport, Oregon. It addresses the problem of resource model identification, that is, the question: Which of various competing models correctly describes the resource? A major theme of the paper is the proposition that data uncertainty and model ambiguity are fundamentally intertwined. In some cases this relationship can be proved rigorously; consequently, model ambiguity can be inevitable. In other cases, promising identification techniques are suggested by modern statistical theory, and the paper examines some of these in the context of time series analysis. Throughout the paper, theoretical and practical examples illustrate both limitations and possibilities for using resource models as valid descriptors of reality. © 1987 Wiley Periodicals, Inc.