How to produce actual Gumbel CDF in Matlab?

Hello,

I aim to fit data samples to a Gumbel distribution (through the CDF) for extreme value analysis. I have seen the 'evfit' to be the only function available in Matlab for Gumbel. The Gumbel CDF's expression is:

P(X) = exp[-exp[-(X-A)/B]] (1)

However, when editing 'evcdf.m' in Matlab (R2012a) the expression that is actually computed is (l.65):

P(X) = 1-exp[-exp[(X-A)/B]] (2)

which is slightly different. How can I make (2) consistent with expression (1) then?

Thanks for your help! JBS.

[José-Luis]

Evaluate at -X, pass -A and do a sum.