SCENARIO PROBABILITY ELICITATION PROCEDURES (DELPHI)
Abstract
Scenario probabilities elicited from experts are useful in decision making under uncertain future conditions, especially when a reliable and sufficient source of past data is unavailable. Various approaches have been proposed, and used, to simplify the determination of scenario probabilities. These include micro level approaches based on the analysis of relevant underlying events and their interrelations, and direct macro level examination of the scenarios. The determination of a unique solution demands excessive consistency and time requirements on the part of the expert and is often not guaranteed by many of these procedures. The procedure described in this thesis is designed to elicit cognitively simple and reliable information from experts in order to determine the probabilities of various possible scenarios. Queries presented to the expert are selected at each step, to maximize the expected information content of the expert's response. A procedure to select the queries which provide the most information, and generate a partial ranking for the scenario probabilities in the presence of inconsistent ranking responses is developed. The services of either a single expert of a panel of judges can be utilized. The Delphi procedure is explained through a model based on catastrophe theory. It is also demonstrated that splitting up the query process for the conditional probability range estimates, and the scenario probability ranks, is desirable when the opinions of a group of experts are obtained. A combined Delphi and filtering approach for determining a group consensus is described. A theoretically valid procedure to break up a large problem, and use separate groups of experts to estimate the scenarios in the different subsets is proposed. A procedure is proposed to evaluate the immunity, or lack of susceptibility, of a decision tree to unforeseen events using some elementary assumptions about the effects of such events and their probabilities. This results in the estimation of a surprise factor which provides a measure of the need for considering the possibility of surprises in reaching an optimal decision. A computer program has been written to implement this information maximizing query procedure (IMQP).
Degree
Ph.D.
Subject Area
Management
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.