Extreme Value Theory Approach to Simultaneous Monitoring and Thresholding of Multiple Risk Indicators

J.H.J. Einmahl, J. Li, R.Y. Liu

Research output: Working paperDiscussion paperOther research output

253 Downloads (Pure)

Abstract

Risk assessments often encounter extreme settings with very few or no occurrences in reality.Inferences about risk indicators in such settings face the problem of insufficient data.Extreme value theory is particularly well suited for handling this type of problems.This paper uses a multivariate extreme value theory approach to establish thresholds for signaling levels of risk in the context of simultaneous monitoring of multiple risk indicators.The proposed threshold system is well justified in terms of extreme multivariate quantiles, and its sample estimator is shown to be consistent.As an illustration, the proposed approach is applied to developing a threshold system for monitoring airline performance measures.This threshold system assigns different risk levels to observed airline performance measures.In particular, it divides the sample space into regions with increasing levels of risk.Moreover, in the univariate case, such a thresholding technique can be used to determine a suitable cut-off point on a runway for holding short of landing aircrafts.This cut-off point is chosen to ensure a certain required level of safety when allowing simultaneous operations on two intersecting runways in order to ease air traffic congestion.
Original languageEnglish
Place of PublicationTilburg
PublisherEconometrics
Number of pages29
Volume2006-104
Publication statusPublished - 2006

Publication series

NameCentER Discussion Paper
Volume2006-104

Keywords

  • Extreme value theory
  • extreme quantile
  • multiple risk indicators
  • multivariate quantile
  • rare event
  • statistics of extremes
  • threshold system

Fingerprint

Dive into the research topics of 'Extreme Value Theory Approach to Simultaneous Monitoring and Thresholding of Multiple Risk Indicators'. Together they form a unique fingerprint.

Cite this