[R] Fit Frechet Distribution

[Ben Bolker]

 <riedwyl <at> giub.unibe.ch> writes:

> 
> hello everybody
> i want to use the maximum likelihood method to estimate FRECHET parameters of 
> my sample data.
> Should it work with fitdistr in the package MASS? 
> I only find how to do it for GEV, Gumbel, and almost all other distributions, 
> but FRECHET?
> I would be very happy if somebody can tell me how to do fit the FRECHET 
> distribution!
> Thanks
> Nadja Riedwyl
> 

the evd package from CRAN has d/r/p/qfrechet functions.

you can set up a negative log-likelihood function

likfun <-  function(loc,shape,scale) {
  -sum(dfrechet(x,loc=loc,shape=shape,scale=scale,log=TRUE))
}

[totally untested]

and use mle, from the stats4 package (or just plain
old optim()) to find the parameters.
The hardest part may be to find good starting parameters.

  Ben Bolker