The Fréchet distribution with shape parameter and scale parameter
has probability
density function and distribution function
|
(1)
| |||
|
(2)
|
for , with
for
. It is implemented in the
Wolfram Language as FrechetDistribution[alpha,
s].
The th
raw moment is
|
(3)
|
for . In particular, the mean,
variance, and median are
|
(4)
| |||
|
(5)
| |||
|
(6)
|
The mean is finite for , and the variance is
finite for
.
The Fréchet distribution is heavy-tailed, with
|
(7)
|
as . It is the type II extreme value distribution
in the Fisher-Tippett-Gnedenko theorem
and, up to location and scale, corresponds to a positive generalized extreme value
shape parameter
. The Pareto
distribution and other distributions with regularly varying upper tails belong
to its maximum domain of attraction (de Haan and Ferreira 2006).
If
has a Fréchet distribution with parameters
and
, then
has a Weibull distribution
with shape parameter
and unit scale.