Median loss decision theory

Chi Wai Yu, Bertrand Clarke

Research output: Contribution to journalReview article

2 Citations (Scopus)

Abstract

In this paper, we argue that replacing the expectation of the loss in statistical decision theory with the median of the loss leads to a viable and useful alternative to conventional risk minimization particularly because it can be used with heavy tailed distributions. We investigate three possible definitions for such medloss estimators and derive examples of them in several standard settings. We argue that the medloss definition based on the posterior distribution is better than the other two definitions that do not permit optimization over large classes of estimators. We argue that median loss minimizing estimates often yield improved performance, have resistance to outliers as high as the usual robust estimates, and are resistant to the specific loss used to form them. In simulations with the posterior medloss formulation, we show how the estimates can be obtained numerically and that they can have better robustness properties than estimates derived from risk minimization.

Original languageEnglish (US)
Pages (from-to)611-623
Number of pages13
JournalJournal of Statistical Planning and Inference
Volume141
Issue number2
DOIs
StatePublished - Feb 2011

Fingerprint

Decision Theory
Decision theory
Statistical Decision Theory
Estimate
Estimator
Robust Estimate
Heavy-tailed Distribution
Posterior distribution
Outlier
Robustness
Median
Optimization
Formulation
Alternatives
Simulation
Risk minimization

Keywords

  • Asymptotics
  • Decision theory
  • Loss function
  • Median
  • Posterior

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty
  • Applied Mathematics

Cite this

Median loss decision theory. / Yu, Chi Wai; Clarke, Bertrand.

In: Journal of Statistical Planning and Inference, Vol. 141, No. 2, 02.2011, p. 611-623.

Research output: Contribution to journalReview article

Yu, Chi Wai ; Clarke, Bertrand. / Median loss decision theory. In: Journal of Statistical Planning and Inference. 2011 ; Vol. 141, No. 2. pp. 611-623.
@article{49a964c253e0457caefba98095242d55,
title = "Median loss decision theory",
abstract = "In this paper, we argue that replacing the expectation of the loss in statistical decision theory with the median of the loss leads to a viable and useful alternative to conventional risk minimization particularly because it can be used with heavy tailed distributions. We investigate three possible definitions for such medloss estimators and derive examples of them in several standard settings. We argue that the medloss definition based on the posterior distribution is better than the other two definitions that do not permit optimization over large classes of estimators. We argue that median loss minimizing estimates often yield improved performance, have resistance to outliers as high as the usual robust estimates, and are resistant to the specific loss used to form them. In simulations with the posterior medloss formulation, we show how the estimates can be obtained numerically and that they can have better robustness properties than estimates derived from risk minimization.",
keywords = "Asymptotics, Decision theory, Loss function, Median, Posterior",
author = "Yu, {Chi Wai} and Bertrand Clarke",
year = "2011",
month = "2",
doi = "10.1016/j.jspi.2010.08.013",
language = "English (US)",
volume = "141",
pages = "611--623",
journal = "Journal of Statistical Planning and Inference",
issn = "0378-3758",
publisher = "Elsevier",
number = "2",

}

TY - JOUR

T1 - Median loss decision theory

AU - Yu, Chi Wai

AU - Clarke, Bertrand

PY - 2011/2

Y1 - 2011/2

N2 - In this paper, we argue that replacing the expectation of the loss in statistical decision theory with the median of the loss leads to a viable and useful alternative to conventional risk minimization particularly because it can be used with heavy tailed distributions. We investigate three possible definitions for such medloss estimators and derive examples of them in several standard settings. We argue that the medloss definition based on the posterior distribution is better than the other two definitions that do not permit optimization over large classes of estimators. We argue that median loss minimizing estimates often yield improved performance, have resistance to outliers as high as the usual robust estimates, and are resistant to the specific loss used to form them. In simulations with the posterior medloss formulation, we show how the estimates can be obtained numerically and that they can have better robustness properties than estimates derived from risk minimization.

AB - In this paper, we argue that replacing the expectation of the loss in statistical decision theory with the median of the loss leads to a viable and useful alternative to conventional risk minimization particularly because it can be used with heavy tailed distributions. We investigate three possible definitions for such medloss estimators and derive examples of them in several standard settings. We argue that the medloss definition based on the posterior distribution is better than the other two definitions that do not permit optimization over large classes of estimators. We argue that median loss minimizing estimates often yield improved performance, have resistance to outliers as high as the usual robust estimates, and are resistant to the specific loss used to form them. In simulations with the posterior medloss formulation, we show how the estimates can be obtained numerically and that they can have better robustness properties than estimates derived from risk minimization.

KW - Asymptotics

KW - Decision theory

KW - Loss function

KW - Median

KW - Posterior

UR - http://www.scopus.com/inward/record.url?scp=77956114004&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77956114004&partnerID=8YFLogxK

U2 - 10.1016/j.jspi.2010.08.013

DO - 10.1016/j.jspi.2010.08.013

M3 - Review article

AN - SCOPUS:77956114004

VL - 141

SP - 611

EP - 623

JO - Journal of Statistical Planning and Inference

JF - Journal of Statistical Planning and Inference

SN - 0378-3758

IS - 2

ER -