Abstract
In economic decision analyses, continuous uncertainties are often represented by discrete probability distributions. In this article, we analyze the ability of discretizations based on the 10th, 50th, and 90th percentiles to match the mean, variance, skewness, and kurtosis of a wide range of distributions in the Johnson distribution system. In addition, we develop new discretization methods that improve upon current practice. Finally, we demonstrate that all of these methods are special cases from a continuum of weightings and show under which conditions each is most appropriate. Our results provide guidelines for the methods’ applications and limits to their usefulness.
Acknowledgments
We thank the Center for Petroleum Asset Risk Management for partially supporting this work. Further, we thank the editor and anonymous reviewers for their helpful comments and feedback.