I have parameters of two Gumbel distributions ($\mu_1, \beta_1)$ and $(\mu_2, \beta_2)$. Since max of 2 Gumbels is a Gumbel, I'd like to compute $\mu_m, \beta_m$, so that:

$Gumbel(\mu_m,\beta_m)$ = $max(Gumbel(\mu_1, \beta_1), Gumbel(\mu_2, \beta_2))$

Any hints how to do this? I've found a solution for the case when $\beta_1 = \beta_2$ in here (point 6, page 105):


but I'm even not sure how this was computed.


Presumably your Gumbel random variables are independent. A Gumbel random variable $X_i$ with location parameter $\alpha_i$ and scale parameter $\beta_i$ has CDF $$ F_i(x) = \exp(-\exp((\alpha_i - x)/\beta_i))$$ Then the CDF for $X = \max(X_1,X_2)$ is $$F_X(x) = \mathbb P(X_1 \le x) \mathbb P(X_2 \le x) = \exp(-\exp((\alpha_1-x)/\beta_1) - \exp((\alpha_2-x)/\beta_2))$$ This is not a Gumbel distribution unless $\beta_1 = \beta_2 = \beta$, in which case it is $\exp(-exp((\alpha - x)/\beta))$ where $\alpha = \beta \ln(\exp(\alpha_1/\beta) + \exp(\alpha_2/\beta))$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.