The largest, or smallest, observation in a sample has one of three possible
distributions. This is another example of convergence in distribution.
The average of n samples taken from any distribution with finite mean and
variance will have a normal distribution for large n. This is the
largest member of a sample of size n has a LEV,
Type I largest extreme value distribution, also called Gumbel, distribution, regardless of the parent
population, IF the parent has an unbounded tail that decreases at least as
fast as an exponential function, and has finite moments (as does the normal,
The LEV, has pdf given by
are the location and scale* parameters, respectively, and
Similar sampling of the smallest member of a
sample of size n produces an SEV, Type I
smallest extreme value
distribution, with density
as n increases. There are two other extreme value distributions. If not all
moments exist for the initial distribution, the largest observation follows a
Type II or Frechet distribution. If the parent density has a bounded tail,
the smallest observation in a sample of size n, has a Type III, or
distribution of minima, as n increases. Examples are smallest samples taken
from lognormal, Gamma, Beta or Weibull distributions.
The Weibull distribution is most easily described by its cdf:
the scale (not location) parameter, and
the shape parameter. Weibull is not a location, scale
Notice that if x has a Weibull distribution, then loge(x) is SEV,
so SEV is to Weibull, as normal is to lognormal. Type I and Type III
limiting distributions are useful in describing physical phenomena where the
outcome is determined by the behavior of the best, or worst, in the sample.
The Weibull Distribution Has Considerable Flexibility.
* A probability density is a location, scale density if it can take the form
where is a proper density and does not depend on any unknown parameters.
is the location parameter and
is the scale parameter. The
Normal, or Gaussian, is a special case with =
= mean, and =
standard deviation. Although location, scale densities are sometimes written
, as generic parameters,
these do not, in general, refer to the
mean and standard deviation of a location, scale density. Note again that
not all densities are location, scale.
A proper density is one for which