Question GEVFIT in fExtremes

[Yohan Chalabi]

>>>> "BJ" == Brommundt Jürgen <juergen.brommundt at colenco.ch>
>>>> on Tue, 20 May 2008 15:16:38 +0200


 BJ> Hello,
 BJ> Doing some extreme value evaluations of wind speed data I
 BJ> found your very useful R-Package fExtremes. Since I started
 BJ> doing my evaluations in MatLab I had a chance to compare the
 BJ> results between the results of the MatLab gevfit and your gevfit
 BJ> routine. Because we need the data for the design of buildings
 BJ> I am interested in extremes of given return intervalls and
 BJ> the uncertainty, estimated from the confidence intervalls.
 BJ> Matlab's gevfit estimates the parameters of the GEV and provides
 BJ> another set of parameters for the limits of the confidence
 BJ> intervalls. From these I estimated the design values using
 BJ> the inverse of the GEV. From fExtremes I used the function
 BJ> gevrlevelPlot with the same confidence intervall (c=0.9)
 BJ> to get the uncertainty bands.
 BJ> As you can see from the attached Excel-File the MLM-fits of
 BJ> the GEV agree very well from both programs. Unfortunately the
 BJ> confidence bands do not agree. I tried to find out the reason
 BJ> from the documentation, but unfortunately did not succeed. I
 BJ> would appreciate, if you could be so kind, to give me a hint,
 BJ> what a possible reason might be??
 BJ> Thanks a lot in advance!
 BJ> 
 BJ> Kind regards,
 BJ> Jürgen Brommundt

Dear Jürgen,

Could you please send me your matlab, R code and data set such that I
can reproduce your example.

sorry for the very late response,
Yohan


-- 
PhD student
Swiss Federal Institute of Technology
Zurich

www.ethz.ch
www.rmetrics.org

NOTE:
Rmetrics Workshop: http://www.rmetrics.org/meielisalp.htm
June 29th - July 3rd Meielisalp, Lake Thune, Switzerland