Jump to content

Toggle sidebar
Search Go

 Create account
 Log in
Personal tools



Photograph your local culture, help Wikipedia and win!

Contents
 hide

(Top)

Data analysis

Applications

History

Univariate theory


Multivariate theory

Nonstationary extremes

See also

Notes

References

Software

External links

Extreme value theory

9 languages
 Article
 Talk
 Read
 Edit
 View history
From Wikipedia, the free encyclopedia
This article is about the statistical theory. For the result in calculus, see extreme value
theorem.

Extreme value theory is used to model the risk of extreme, rare events, such as the 1755 Lisbon earthquake.

Extreme value theory or extreme value analysis (EVA) is a branch

of statistics dealing with the extreme deviations from the median of probability
distributions. It seeks to assess, from a given ordered sample of a given random
variable, the probability of events that are more extreme than any previously observed.
Extreme value analysis is widely used in many disciplines, such as structural
engineering, finance, earth sciences, traffic prediction, and geological engineering. For
example, EVA might be used in the field of hydrology to estimate the probability of an
unusually large flooding event, such as the 100-year flood. Similarly, for the design of
a breakwater, a coastal engineer would seek to estimate the 50-year wave and design
the structure accordingly.

Data analysis[edit]
Two main approaches exist for practical extreme value analysis.
The first method relies on deriving block maxima (minima) series as a preliminary step.
In many situations it is customary and convenient to extract the annual maxima
(minima), generating an "Annual Maxima Series" (AMS).
The second method relies on extracting, from a continuous record, the peak values
reached for any period during which values exceed a certain threshold (falls below a
certain threshold). This method is generally referred to as the "Peak Over
Threshold"[1] method (POT).
For AMS data, the analysis may partly rely on the results of the Fisher–Tippett–
Gnedenko theorem, leading to the generalized extreme value distribution being selected
for fitting.[2][3] However, in practice, various procedures are applied to select between a
wider range of distributions. The theorem here relates to the limiting distributions for the
minimum or the maximum of a very large collection of independent random
variables from the same distribution. Given that the number of relevant random events
within a year may be rather limited, it is unsurprising that analyses of observed AMS
data often lead to distributions other than the generalized extreme value distribution
(GEVD) being selected.[4]
For POT data, the analysis may involve fitting two distributions: one for the number of
events in a time period considered and a second for the size of the exceedances.
A common assumption for the first is the Poisson distribution, with the generalized
Pareto distribution being used for the exceedances. A tail-fitting can be based on
the Pickands–Balkema–de Haan theorem.[5][6]
Novak[7] reserves the term “POT method” to the case where the threshold is non-
random, and distinguishes it from the case where one deals with exceedances of a
random threshold.

Applications[edit]
Applications of extreme value theory include predicting the probability distribution of:

 Extreme floods; the size of freak waves

 Tornado outbreaks[8]
 Maximum sizes of ecological populations[9]
 Side effects of drugs (e.g., ximelagatran)
 The magnitudes of large insurance losses
 Equity risks; day-to-day market risk
 Mutational events during evolution
 Large wildfires[10]
 Environmental loads on structures[11]
 Fastest time humans are capable of running the 100 metres sprint[12] and
performances in other athletic disciplines[13][14][15]
 Pipeline failures due to pitting corrosion
 Anomalous IT network traffic, prevent attackers from reaching important data
 Road safety analysis[16][17]
 Wireless communications[18]
 Epidemics[19]
 Neurobiology[20]

History[edit]
The field of extreme value theory was pioneered by Leonard Tippett (1902–1985).
Tippett was employed by the British Cotton Industry Research Association, where he
worked to make cotton thread stronger. In his studies, he realized that the strength of a
thread was controlled by the strength of its weakest fibres. With the help of R. A. Fisher,
Tippet obtained three asymptotic limits describing the distributions of extremes
assuming independent variables. Emil Julius Gumbel codified this theory in his 1958
book Statistics of Extremes, including the Gumbel distributions that bear his name.
These results can be extended to allow for slight correlations between variables, but the
classical theory does not extend to strong correlations of the order of the variance. One
universality class of particular interest is that of log-correlated fields, where the
correlations decay logarithmically with the distance.
Univariate theory[edit]
Let  be a sequence of independent and identically distributed random variables
with cumulative distribution function F and let  denote the maximum.
In theory, the exact distribution of the maximum can be derived:
The associated indicator function  is a Bernoulli process with a success
probability  that depends on the magnitude  of the extreme event. The number of
extreme events within  trials thus follows a binomial distribution and the number of
trials until an event occurs follows a geometric distribution with expected value and
standard deviation of the same order .
In practice, we might not have the distribution function  but the Fisher–Tippett–
Gnedenko theorem provides an asymptotic result. If there exist sequences of
constants  and  such that
as  then
where  depends on the tail shape of the distribution. When
normalized, G belongs to one of the following non-degenerate
distribution families:
Weibull law:  when the distribution of  has a light tail with finite upper bound.
Also known as Type 3.
Gumbel law:  when the distribution of  has an exponential tail. Also known as
Type 1.
Fréchet law:  when the distribution of  has a heavy tail (including polynomial
decay). Also known as Type 2.
For the Weibull and Fréchet laws, .

Multivariate theory[edit]
Extreme value theory in more than one variable introduces additional issues
that have to be addressed. One problem that arises is that one must specify
what constitutes an extreme event.[21] Although this is straightforward in the
univariate case, there is no unambiguous way to do this in the multivariate
case. The fundamental problem is that although it is possible to order a set of
real-valued numbers, there is no natural way to order a set of vectors.
As an example, in the univariate case, given a set of observations  it is
straightforward to find the most extreme event simply by taking the maximum
(or minimum) of the observations. However, in the bivariate case, given a set
of observations , it is not immediately clear how to find the most extreme
event. Suppose that one has measured the values  at a specific time and the
values  at a later time. Which of these events would be considered more
extreme? There is no universal answer to this question.
Another issue in the multivariate case is that the limiting model is not as fully
prescribed as in the univariate case. In the univariate case, the model (GEV
distribution) contains three parameters whose values are not predicted by the
theory and must be obtained by fitting the distribution to the data. In the
multivariate case, the model not only contains unknown parameters, but also
a function whose exact form is not prescribed by the theory. However, this
function must obey certain constraints.[22][23] It is not straightforward to devise
estimators that obey such constraints though some have been recently
constructed.[24] [25] [26]
As an example of an application, bivariate extreme value theory has been
applied to ocean research.[21][27]

Nonstationary extremes[edit]
Statistical modeling for nonstationary time series was developed in the
1990's.[28] Methods for nonstationary multivariate extremes have been
introduced more recently.[29] The latter can be used for tracking how the
dependence between extreme values changes over time, or over another
covariate.[30][31][32]

See also[edit]
 Extreme risk
 Extreme weather
 Fisher–Tippett–Gnedenko theorem
 Generalized extreme value distribution
 Large deviation theory
 Outlier
 Pareto distribution
 Pickands–Balkema–de Haan theorem
 Rare events
 Weibull distribution
 Redundancy principle
This article includes a list of general references, but it lacks
sufficient corresponding inline citations. Please help
to improve this article by introducing more precise
citations. (September 2010)  (Learn how and when to remove this template
message)

Notes[edit]
1. ^ Leadbetter, M. R. (1991). "On a basis for 'Peaks over Threshold' modeling". Statistics
and Probability Letters.  12  (4): 357–362.  doi:10.1016/0167-7152(91)90107-3.
2. ^ Fisher and Tippett (1928)
3. ^ Gnedenko (1943)
4. ^ Embrechts, Klüppelberg, and Mikosch (1997)
5. ^ Pickands (1975)
6. ^ Balkema and de Haan (1974)
7. ^ Novak (2011)
8. ^ Tippett, Michael K.; Lepore, Chiara; Cohen, Joel E. (16 December 2016). "More
tornadoes in the most extreme U.S. tornado outbreaks".  Science.  354  (6318): 1419–
1423.  Bibcode:2016Sci...354.1419T.  doi:10.1126/science.aah7393. PMID 27934705.
9. ^ Batt, Ryan D.; Carpenter, Stephen R.; Ives, Anthony R. (March 2017). "Extreme
events in lake ecosystem time series". Limnology and Oceanography Letters.  2 (3):
63.  doi:10.1002/lol2.10037.
10. ^ Alvardo (1998, p.68.)
11. ^ Makkonen (2008)
12. ^ J.H.J. Einmahl & S.G.W.R. Smeets (2009),  "Ultimate 100m World Records Through
Extreme-Value Theory"  (PDF),  CentER Discussion Paper, Tilburg University, 57,
archived from the original  (PDF) on 2016-03-12, retrieved 2009-08-12
13. ^ D. Gembris, J.Taylor & D. Suter (2002), "Trends and random fluctuations in
athletics", Nature, 417 (6888):
506,  Bibcode:2002Natur.417..506G,  doi:10.1038/417506a,  hdl:2003/25362, PMID 120
37557, S2CID  13469470
14. ^ D. Gembris, J.Taylor & D. Suter (2007), "Evolution of athletic records : Statistical
effects versus real improvements", Journal of Applied Statistics, 34 (5): 529–
545,  doi:10.1080/02664760701234850, hdl:2003/25404,  S2CID 55378036
15. ^ H. Spearing, J. Tawn, D. Irons, T. Paulden & G. Bennett (2021), "Ranking, and other
properties, of elite swimmers using extreme value theory", Journal of the Royal
Statistical Society: Series A (Statistics in Society),  184  (1): 368–
395,  doi:10.1111/rssa.12628, S2CID  204823947
16. ^ Songchitruksa, P.; Tarko, A. P. (2006). "The extreme value theory approach to safety
estimation".  Accident Analysis and Prevention.  38  (4): 811–
822.  doi:10.1016/j.aap.2006.02.003.  PMID  16546103.
17. ^ Orsini, F.; Gecchele, G.; Gastaldi, M.; Rossi, R. (2019). "Collision prediction in
roundabouts: a comparative study of extreme value theory
approaches". Transportmetrica A: Transport Science.  15  (2): 556–
572.  doi:10.1080/23249935.2018.1515271.  S2CID 158343873.
18. ^ C. G. Tsinos, F. Foukalas, T. Khattab and L. Lai, "On Channel Selection for Carrier
Aggregation Systems." IEEE Transactions on Communications, vol. 66, no. 2, Feb. 2018
) 808-818.
19. ^ Wong, Felix; Collins, James J. (2020-11-02). "Evidence that coronavirus
superspreading is fat-tailed".  Proceedings of the National Academy of
Sciences.  117  (47): 29416–
29418. Bibcode:2020PNAS..11729416W. doi:10.1073/pnas.2018490117. ISSN 0027-
8424.  PMC 7703634. PMID 33139561.
20. ^ Basnayake, Kanishka; Mazaud, David; Bemelmans, Alexis; Rouach, Nathalie;
Korkotian, Eduard; Holcman, David (2019-06-04).  "Fast calcium transients in dendritic
spines driven by extreme statistics".  PLOS Biology. 17 (6):
e2006202.  doi:10.1371/journal.pbio.2006202.  ISSN  1545-7885. PMC  6548358.  PMID 
31163024.
21. ^ Jump up to:a b Morton, I.D.; Bowers, J. (December 1996). "Extreme value analysis in a
multivariate offshore environment".  Applied Ocean Research.  18  (6): 303–
317.  doi:10.1016/s0141-1187(97)00007-2.  ISSN  0141-1187.
22. ^ Beirlant, Jan; Goegebeur, Yuri; Teugels, Jozef; Segers, Johan (2004-08-
27).  Statistics of Extremes: Theory and Applications. Wiley Series in Probability and
Statistics. Chichester, UK: John Wiley & Sons,
Ltd. doi:10.1002/0470012382.  ISBN  9780470012383.
23. ^ Coles, Stuart (2001). An Introduction to Statistical Modeling of Extreme Values.
Springer Series in Statistics.  doi:10.1007/978-1-4471-3675-0.  ISBN  978-1-84996-874-
4. ISSN 0172-7397.
 24. ^ de Carvalho, M.; Davison, A. C. (2014).  "Spectral density ratio models for multivariate
extremes"  (PDF). Journal of the American Statistical Association.  109: 764‒
776.  doi:10.1016/j.spl.2017.03.030.
25. ^ Hanson, T.; de Carvalho, M.; Chen, Yuhui (2017). "Bernstein polynomial angular
densities of multivariate extreme value distributions"  (PDF).  Statistics and Probability
Letters.  128: 60–66. doi:10.1016/j.spl.2017.03.030.
26. ^ de Carvalho, M. (2013).  "A Euclidean likelihood estimator for bivariate tail
dependence"  (PDF).  Communications in Statistics – Theory and Methods. 42 (7):
1176–1192.  arXiv:1204.3524.  doi:10.1080/03610926.2012.709905.  S2CID 42652601.
27. ^ Zachary, S.; Feld, G.; Ward, G.; Wolfram, J. (October 1998). "Multivariate
extrapolation in the offshore environment".  Applied Ocean Research.  20  (5): 273–
295.  doi:10.1016/s0141-1187(98)00027-3.  ISSN  0141-1187.
28. ^ Davison, A.C.; Smith, Richard (1990).  "Models for exceedances over high
thresholds". Journal of the Royal Statistical Society: Series B (Methodological).  52  (3):
393–425. doi:10.1111/j.2517-6161.1990.tb01796.x.
29. ^ de Carvalho, M. (2016).  Statistics of extremes: Challenges and opportunities. In:
Handbook of EVT and its Applications to Finance and Insurance  (PDF). Hoboken:
Wiley. p. 195--214.  ISBN  978-1-118-65019-6.
30. ^ Castro, D.; de Carvalho, M.; Wadsworth, J. (2018). "Time-Varying Extreme Value
Dependence with Application to Leading European Stock Markets"  (PDF).  Annals of
Applied Statistics.  12: 283–309.  doi:10.1214/17-AOAS1089.  S2CID 33350408.
31. ^ Mhalla, L.; de Carvalho, M.; Chavez-Demoulin, V. (2019). "Regression Type Models
for Extremal Dependence"  (PDF).  Scandinavian Journal of Statistics. 46 (4): 1141–
1167.  doi:10.1111/sjos.12388.  S2CID 53570822.
32. ^ Mhalla, L.; de Carvalho, M.; Chavez-Demoulin, V. (2018). "Local robust estimation of
the Pickands dependence function".  Annals of Statistics. 46 (6A): 2806–
2843.  doi:10.1214/17-AOS1640. S2CID  59467614.

References[edit]
 Abarbanel, H.; Koonin, S.; Levine, H.; MacDonald, G.; Rothaus, O.
(January 1992), "Statistics of Extreme Events with Application to
Climate"  (PDF), JASON, JSR-90-30S, retrieved 2015-03-03
 Alvarado, Ernesto; Sandberg, David V.; Pickford, Stewart G.
(1998), "Modeling Large Forest Fires as Extreme
Events"  (PDF), Northwest Science, 72: 66–75, archived from the
original  (PDF) on 2009-02-26, retrieved 2009-02-06
 Balkema, A.; Laurens (1974), "Residual life time at great age", Annals of
Probability, 2 (5): 792–
804, doi:10.1214/aop/1176996548, JSTOR 2959306
 Burry K.V. (1975). Statistical Methods in Applied Science. John Wiley &
Sons.
 Castillo E. (1988) Extreme value theory in engineering. Academic Press,
Inc. New York. ISBN 0-12-163475-2.
 Castillo, E., Hadi, A. S., Balakrishnan, N. and Sarabia, J. M. (2005)
Extreme Value and Related Models with Applications in Engineering and
Science, Wiley Series in Probability and Statistics Wiley, Hoboken, New
Jersey. ISBN 0-471-67172-X.
 Coles S. (2001) An Introduction to Statistical Modeling of Extreme
Values. Springer, London.
 Embrechts P., Klüppelberg C. and Mikosch T. (1997) Modelling extremal
events for insurance and finance. Berlin: Spring Verlag
 Fisher, R.A.; Tippett, L.H.C. (1928), "Limiting forms of the frequency
distribution of the largest and smallest member of a sample", Proc.
Camb. Phil. Soc., 24 (2): 180–
190, Bibcode:1928PCPS...24..180F, doi:10.1017/s0305004100015681, 
S2CID 123125823
 Gnedenko, B.V. (1943), "Sur la distribution limite du terme maximum
d'une serie aleatoire", Annals of Mathematics, 44 (3): 423–
453, doi:10.2307/1968974, JSTOR 1968974
 Gumbel, E.J. (1935), "Les valeurs extrêmes des distributions
statistiques"  (PDF), Annales de l'Institut Henri Poincaré, 5 (2): 115–158,
retrieved 2009-04-01
 Gumbel, E. J. (2004) [1958], Statistics of Extremes, Mineola, NY:
Dover, ISBN 978-0-486-43604-3
 Makkonen, L. (2008), "Problems in the extreme value
analysis", Structural Safety, 30 (5): 405–
419, doi:10.1016/j.strusafe.2006.12.001
 Leadbetter, M. R. (1991), "On a basis for 'Peaks over Threshold'
modeling", Statistics & Probability Letters, 12 (4): 357–
362, doi:10.1016/0167-7152(91)90107-3
 Leadbetter M.R., Lindgren G. and Rootzen H. (1982) Extremes and
related properties of random sequences and processes. Springer-Verlag,
New York.
 Lindgren, G.; Rootzen, H. (1987), "Extreme values: Theory and technical
applications", Scandinavian Journal of Statistics, Theory and
Applications, 14: 241–279
 Novak S.Y. (2011) Extreme Value Methods with Applications to Finance.
Chapman & Hall/CRC Press, London. ISBN 978-1-4398-3574-6
 Pickands, J (1975), "Statistical inference using extreme order
statistics", Annals of Statistics, 3: 119–
131, doi:10.1214/aos/1176343003

Software[edit]
 Extreme Value Statistics in R - Packages for extreme value statistics in R
 ExtremeStats.jl and Extremes.jl - Extreme Value Statistics in Julia

External links[edit]
 Extreme Value Theory can save your neck Easy non-mathematical
introduction (pdf)
 Source Code for Stationary and Nonstationary Extreme Value
Analysis University of California, Irvine
  Steps in Applying Extreme Value Theory to Finance: A Review
 Les valeurs extrêmes des distributions statistiques Full-text access to
conferences held by E. J. Gumbel in 1933–34, in French (pdf)
Categories: 
 Actuarial science
 Statistical theory
 Extreme value data
 Tails of probability distributions
 Financial risk modeling
 This page was last edited on 9 February 2023, at 09:32 (UTC).
 Text is available under the Creative Commons Attribution-ShareAlike License 3.0; additional terms
may apply. By using this site, you agree to the Terms of Use and Privacy Policy. Wikipedia® is a
registered trademark of the Wikimedia Foundation, Inc., a non-profit organization.
 Privacy policy

 About Wikipedia

 Disclaimers

 Contact Wikipedia

 Mobile view

 Developers

 Statistics

 Cookie statement


In probability theory and statistics, a probability distribution is the mathematical
function that gives the probabilities of occurrence of different
possible outcomes for an experiment. It is a mathematical description of a random
phenomenon in terms of its sample space and the probabilities of events.