Discretization in decision analysis

Date

2020-08

Authors

Woodruff, Joshua Morgan

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

The choice of discretizations in Decision Analysis impacts the accuracy of the probabilistic analysis of the potential strategies. This dissertation introduces a novel method for creating discretizations for specific problems. Next, we introduce the distance metric, which is borrowed from stochastic optimization. This metric indicates how well two cumulative distribution functions match each other in terms of shape. Discretizations that better match the shape are more accurate in estimating the value of a cumulative distribution function at any given percentile. We determine under which conditions the distance is higher or lower, and which discretizations to choose. Finally, we show what happens to the accuracy of discretizations when there is assessment error and how this impacts the choice of discretizations.

Description

LCSH Subject Headings

Citation