I want to fit a Frechet distribution to the tail of the empirical loss distribution using Matlab. However Matlab does not have Frechet distribution. The more general distribution which includes also Frechet is the generalized Pareto distribution, but Matlab does not have this one eather. It has only Weibull and Extreme Value distributions (I think the available Extreme Value distribution is equal to the log of the Weibull distribution).Do you have some ideas how would I solve this problem?I will also need to canculate the inverse of the distribution for some confidence interval.

I think that extreme value distributions include: Frechet, Gumbel and Weibull (?)

I would think so too. but when you read a description of which extreme value distribution is available in Matlab, then I guess you would probably think differently.I know, that the generalised Pareto distribution includes Weibull, Gumbel and Frechet, but the one available in Matlab (extreme value distribution) is equal to log(Weibull).But I have found a software on Internet "EVIM", which is on EVT analysis. I am reading now and I hope that this is something that I need.I need to finish this project for monday, so I hope that EVIM will help to solve the tail fitting problem.

well, maybe the extreme value distribution which is in Matlab, is indeed the generalised Pareto distribution, but written in different way that I have seen in literature. Normally one has a shape parameter /ksi which controles the shape of the distribution. I wonder then what in the parameter /ksi equevelent to of the Matlab' s version of the EV distribution?