A way to solve this problem is transform the three parameters from the GEV-distribution into Frechet. Have a look at Thomas & Reiss, Statistical Analysis of Extreme Values, or the thesis of Han Zhongxian, Actuarial modelling of extremal events... The last is free in the web. I am myself using the package evd by Alec Stephenson. Best wishes Ekkehardt
Fit Frechet Distribution
1 message · Peters, Ekkehardt