Generalized extreme value distribution

Notation	$\textrm{GEV}(\mu,\,\sigma,\,\xi)$
Parameters	μ ∈ R — location, σ > 0 — scale, ξ ∈ R — shape.
Support	x ∈ [ μ − σ / ξ, +∞)   when ξ > 0, x ∈ (−∞, +∞)   when ξ = 0, x ∈ (−∞, μ − σ / ξ ]   when ξ < 0.
pdf	$\frac{1}{\sigma}\,t(x)^{\xi+1}e^{-t(x)},$   where $t(x) = \begin{cases}\big(1+(\tfrac{x-\mu}{\sigma})\xi\big)^{-1/\xi} & \textrm{if}\ \xi\neq0 \\ e^{-(x-\mu)/\sigma} & \textrm{if}\ \xi=0\end{cases}$
CDF	$e^{-t(x)},\,$   for x ∈ support
Mean	$\begin{cases}\mu + \sigma\frac{\Gamma(1-\xi)-1}{\xi} & \text{if}\ \xi\neq 0,\xi<1,\\ \mu + \sigma\,\gamma & \text{if}\ \xi=0,\\ \infty & \text{if}\ \xi\geq 1,\end{cases}$ where $\gamma$ is Euler’s constant.
Median	$\begin{cases}\mu + \sigma \frac{(\ln2)^{-\xi}-1}{\xi} & \text{if}\ \xi\neq0,\\ \mu - \sigma \ln\ln2 & \text{if}\ \xi=0.\end{cases}$
Mode	$\begin{cases}\mu + \sigma \frac{(1+\xi)^{-\xi}-1}{\xi} & \text{if}\ \xi\neq0,\\ \mu & \text{if}\ \xi=0.\end{cases}$
Variance	$\begin{cases}\sigma^2\,(g_2-g_1^2)/\xi^2 & \text{if}\ \xi\neq0,\xi<\frac12,\\ \sigma^2\,\frac{\pi^2}{6} & \text{if}\ \xi=0, \\ \infty & \text{if}\ \xi\geq\frac12,\end{cases}$ where gk = Γ(1 − kξ).
Skewness	$\begin{cases}\frac{g_3-3g_1g_2+2g_1^3}{(g_2-g_1^2)^{3/2}} & \text{if}\ \xi\neq0,\\ \frac{12 \sqrt{6} \zeta(3)}{\pi^3} & \text{if}\ \xi=0.\end{cases}$ where $\zeta(x)$ is Riemann zeta function
Ex. kurtosis	$\begin{cases}\frac{g_4-4g_1g_3+6g_2g_1^2-3g_1^4}{(g_2-g_1^2)^{2}}-3 & \text{if}\ \xi\neq0,\\ \frac{12}{5} & \text{if}\ \xi=0.\end{cases}$
Entropy	$\log(\sigma)\,+\,\gamma\xi\,+\,(\gamma+1)$
MGF
CF

In probability theory and statistics, the generalized extreme value (GEV) distribution is a family of continuous probability distributions developed within extreme value theory to combine the Gumbel, Fréchet and Weibull families also known as type I, II and III extreme value distributions. By the extreme value theorem the GEV distribution is the limit distribution of properly normalized maxima of a sequence of independent and identically distributed random variables. Because of this, the GEV distribution is used as an approximation to model the maxima of long (finite) sequences of random variables.

In some fields of application the generalized extreme value distribution is known as the Fisher–Tippett distribution, named after R. A. Fisher and L. H. C. Tippett who recognised three function forms outlined below. However usage of this name is sometimes restricted to mean the special case of the Gumbel distribution.

Specification

The generalized extreme value distribution has cumulative distribution function

$F(x;\mu,\sigma,\xi) = \exp\left\{-\left[1+\xi\left(\frac{x-\mu}{\sigma}\right)\right]^{-1/\xi}\right\}$

for $1+\xi(x-\mu)/\sigma>0$, where $\mu\in\mathbb R$ is the location parameter, $\sigma>0$ the scale parameter and $\xi\in\mathbb R$ the shape parameter. For $\xi = 0$ the expression is formally undefined and is understood as a limiting case.

The density function is, consequently

$f(x;\mu,\sigma,\xi) = \frac{1}{\sigma}\left[1+\xi\left(\frac{x-\mu}{\sigma}\right)\right]^{(-1/\xi)-1}$$\exp\left\{-\left[1+\xi\left(\frac{x-\mu}{\sigma}\right)\right]^{-1/\xi}\right\}$

again, for $1+\xi(x-\mu)/\sigma>0$.

Summary statistics

Some simple statistics of the distribution are:^{[citation needed]}

$\operatorname{E}(X) = \mu-\frac{\sigma}{\xi}+\frac{\sigma}{\xi}g_1 ,$

$\operatorname{Var}(X) = \frac{\sigma^2}{\xi^2}(g_2-g_1^2) ,$

$\operatorname{Mode}(X) = \mu+\frac{\sigma}{\xi}[(1+\xi)^{-\xi}-1] .$

The skewness is for ξ>0

$\operatorname{skewness}(X) = \frac{g_3-3g_1g_2+2g_1^3}{(g_2-g_1^2)^{3/2}}$

For ξ<0, the sign of the numerator is reversed.

The excess kurtosis is:

$\operatorname{kurtosis\ excess}(X) = \frac{g_4-4g_1g_3+6g_2g_1^2-3g_1^4}{(g_2-g_1^2)^{2}}-3 .$

where $g_k=\Gamma(1+k\xi)$, k=1,2,3,4, and $\Gamma(t)$ is the gamma function.

Link to Fréchet, Weibull and Gumbel families

The shape parameter $\xi$ governs the tail behaviour of the distribution. The sub-families defined by $\xi= 0$, $\xi>0$ and $\xi<0$ correspond, respectively, to the Gumbel, Fréchet and Weibull families, whose cumulative distribution functions are displayed below.

• Gumbel or type I extreme value distribution ($\xi=0$)

$F(x;\mu,\sigma,0)=e^{-e^{-(x-\mu)/\sigma}}\;\;\; \text{for} \;\; x\in\mathbb R.$

• Fréchet or type II extreme value distribution, if $\xi=\alpha^{-1}>0$

$F(x;\mu,\sigma,\xi)=\begin{cases} 0 & x\leq \mu \\ e^{-((x-\mu)/\sigma)^{-\alpha}} & x>\mu. \end{cases}$

• Reversed Weibull or type III extreme value distribution, if $\xi=-\alpha^{-1}<0$

$F(x;\mu,\sigma,\xi)=\begin{cases} e^{-(-(x-\mu)/\sigma)^{\alpha}} & x<\mu \\ 1 & x\geq \mu \end{cases}$

where $\sigma>0$.

Remark I: The theory here relates to maxima and the distribution being discussed is an extreme value distribution for maxima. A generalised extreme value distribution for minima can be obtained, for example by substituting (−x) for x in the distribution function, and subtracting from one: this yields a separate family of distributions.

Remark II: The ordinary Weibull distribution arises in reliability applications and is obtained from the distribution here by using the variable $t = \mu - x$, which gives a strictly positive support - in contrast to the use in the extreme value theory here. This arises because the Weibull distribution is used in cases that deal with the minimum rather than the maximum. The distribution here has an addition parameter compared to the usual form of the Weibull distribution and, in addition, is reversed so that the distribution has an upper bound rather than a lower bound. Importantly, in applications of the GEV, the upper bound is unknown and so must be estimated while when applying the Weibull distribution the lower bound is known to be zero.

Remark III: Note the differences in the ranges of interest for the three extreme value distributions: Gumbel is unlimited, Fréchet has a lower limit, while the reversed Weibull has an upper limit.

One can link the type I to types II and III the following way: if the cumulative distribution function of some random variable $X$ is of type II, and with the positive numbers as support, i.e. $F(x; 0, \sigma, \alpha)$, then the cumulative distribution function of $\ln X$ is of type I, namely $F(x; \ln \sigma, 1/\alpha, 0)$. Similarly, if the cumulative distribution function of $X$ is of type III, and with the negative numbers as support, i.e. $F(x; 0, \sigma, -\alpha)$, then the cumulative distribution function of $\ln (-X)$ is of type I, namely $F(x; -\ln \sigma, 1/\alpha, 0)$.

Link to logit models (logistic regression)

Multinomial logit models, and certain other types of logistic regression, can be phrased as latent variable models with error variables distributed as Gumbel distributions (type I generalized extreme value distributions). This phrasing is common in the theory of discrete choice models, which include logit models, probit models, and various extensions of them, and derives from the fact that the difference of two type-I GEV-distributed variables follows a logistic distribution, of which the logit function is the quantile function. The type-I GEV distribution thus plays the same role in these logit models as the normal distribution does in the corresponding probit models.

Properties

The cumulative distribution function of the generalized extreme value distribution solves the stability postulate equation.^{[citation needed]} The generalized extreme value distribution is a special case of a max-stable distribution, and is a transformation of a min-stable distribution.

Related distributions

• If $X \sim \textrm{GEV}(\mu,\,\sigma,\,0)$ then $mX+b \sim \textrm{GEV}(m\mu+b,\,m\sigma,\,0)$
• If $X \sim \textrm{Gumbel}(\mu,\,\sigma)$ (Gumbel distribution) then $X \sim \textrm{GEV}(\mu,\,\sigma,\,0)$
• If $X \sim \textrm{Weibull}(\mu,\,\sigma)$ (Weibull distribution) then $\mu\left(1-\sigma\mathrm{log}{\tfrac{X}{\sigma}}\right) \sim \textrm{GEV}(\mu,\,\sigma,\,0)$
• If $X \sim \textrm{GEV}(\mu,\,\sigma,\,0)$ then $\sigma \exp (-\tfrac{X-\mu}{\mu \sigma} ) \sim \textrm{Weibull}(\mu,\,\sigma)$ (Weibull distribution)^{[citation needed]}
• If $X \sim \textrm{Exponential}(1)\,$ (Exponential distribution) then $\mu - \sigma \log{X} \sim \textrm{GEV}(\mu,\,\sigma,\,0)$^{[citation needed]}
• If $X \sim \mathrm{GEV}(\alpha,\beta,0)\,$ and $Y \sim \mathrm{GEV}(\alpha,\beta,0)\,$ then $X-Y \sim \mathrm{Logistic}(0,\beta) \,$ (Logistic distribution)
• If $X \sim \mathrm{GEV}(\alpha,\beta,0)\,$ and $Y \sim \mathrm{GEV}(\alpha,\beta,0)\,$ then $X+Y \sim \mathrm{Logistic}(2 \alpha,\beta) \,$

See also

• Fisher–Tippett–Gnedenko theorem
• Generalized Pareto distribution

References

• Embrechts, Paul; Klüppelberg, Claudia; Mikosch, Thomas (1997). Modelling extremal events for insurance and finance. Berlin: Springer Verlag. 
• Leadbetter, M.R., Lindgren, G. and Rootzén, H. (1983). Extremes and related properties of random sequences and processes. Springer-Verlag. ISBN 0-387-90731-9. 
• Resnick, S.I. (1987). Extreme values, regular variation and point processes. Springer-Verlag. ISBN 0-387-96481-9. 
• Coles, Stuart (2001). An Introduction to Statistical Modeling of Extreme Values,. Springer-Verlag. ISBN 1-85233-459-2.