30.09.2014 Views

Package 'extRemes' - What are R and CRAN?

Package 'extRemes' - What are R and CRAN?

Package 'extRemes' - What are R and CRAN?

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

24 devd<br />

Value<br />

independently noted later by meteorologist A. F. Jenkins (1955). It enjoys theretical support for<br />

modeling maxima taken over large blocks of a series of data.<br />

The generalized P<strong>are</strong>o df is given by (Pick<strong>and</strong>s, 1975)<br />

PrX 0, scale > 0, <strong>and</strong> x > threshold. If shape = 0, then the GP<br />

df is defined by continuity <strong>and</strong> becomes<br />

F(x) = 1 - exp(-(x - threshold)/scale).<br />

There is an approximate relationship between the GEV <strong>and</strong> GP df’s where the GP df is approximately<br />

the tail df for the GEV df. In particular, the scale parameter of the GP is a function of the<br />

threshold (denote it scale.u), <strong>and</strong> is equivalent to scale + shape*(threshold - location) where scale,<br />

shape <strong>and</strong> location <strong>are</strong> parameters from the “equivalent” GE Vdf. Similar to the GEV df, the shape<br />

parameter determines the tail behavior, where shape = 0 gives rise to the exponential df (light tail),<br />

shape > 0 the P<strong>are</strong>to df (heavy tail) <strong>and</strong> shape < 0 the Beta df (bounded upper tail at location -<br />

scale.u/shape). Theoretical justification supports the use of the GP df family for modeling excesses<br />

over a high threshold (i.e., y = x - threshold). It is assumed here that x, q describe x (not y = x -<br />

threshold). Similarly, the r<strong>and</strong>om draws <strong>are</strong> y + threshold.<br />

See Coles (2001) <strong>and</strong> Reiss <strong>and</strong> Thomas (2007) for a very accessible text on extreme value analysis<br />

<strong>and</strong> for more theoretical texts, see for example, Beirlant et al. (2004), de Haan <strong>and</strong> Ferreira (2006),<br />

as well as Reiss <strong>and</strong> Thomas (2007).<br />

’devd’ gives the density function, ’pevd’ gives the distribution function, ’qevd’ gives the quantile<br />

function, <strong>and</strong> ’revd’ generates r<strong>and</strong>om deviates for the GEV or GP df depending on the type argument.<br />

Note<br />

There is a similarity between the location parameter of the GEV df <strong>and</strong> the threshold for the GP df.<br />

For clarity, two separate arguments <strong>are</strong> emplyed here to distinguish the two instead of, for example,<br />

just using the location parameter to describe both.<br />

Author(s)<br />

Eric Gillel<strong>and</strong><br />

References<br />

Beirlant, J., Goegebeur, Y., Teugels, J. <strong>and</strong> Segers, J. (2004) Statistics of Extremes: Theory <strong>and</strong><br />

Applications. Chichester, West Sussex, Engl<strong>and</strong>, UK: Wiley, ISBN 9780471976479, 522pp.<br />

Coles, S. (2001) An introduction to statistical modeling of extreme values, London, U.K.: Springer-<br />

Verlag, 208 pp.<br />

de Haan, L. <strong>and</strong> Ferreira, A. (2006) Extreme Value Theory: An Introduction. New York, NY, USA:<br />

Springer, 288pp.<br />

Jenkinson, A. F. (1955) The frequency distribution of the annual maximum (or minimum) of meteorological<br />

elements. Quart. J. R. Met. Soc., 81, 158–171.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!