Proper scoring rules : properties and applications

Proper and strictly proper scoring rules provide a rigorous method for evaluating the accuracy of a probabilistic forecast while encouraging honesty. In this dissertation, we develop new proper and strictly proper scoring rules. We introduce additive and strongly additive scoring rules that can be used to reward a sequence of probabilistic forecasts. We construct new tailored scoring rules and demonstrate a general economic interpretation for all weighted proper scores. We also present a matrix-based construction method for scoring forecasts that can be represented as affine transformations of an underlying distribution.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Operations Research and Industrial Engineering

Grantor

The University of Texas at Austin

Year dc:date.issued

2020

Author dc:creator

• Smith, Zachary James

Advisor dc:contributor.advisor

• Bickel, J. Eric

Committee members dc:contributor.committeemember

• Hasenbein, John J
• Leibowicz, Benjamin D
• Landsberger, Sheldon

• Proper scoring rules
• Generalized entropy

Language dc:language.iso

en

Identifier URI

http://dx.doi.org/10.26153/tsw/13417

OAI identifier oai:identifier

oai:repositories.lib.utexas.edu:2152/86466