2
$\begingroup$

Suppose you have a general random Gaussian vector $\mathbf{X}\sim\mathcal{N}\left(\boldsymbol{\mu},\boldsymbol{\Sigma}\right)$. I'm looking for the simple way to calculate the distribution of the maximal component: $P\left(\mathrm{argmax}_{i}X_{i}=k\right)$. Is there some closed-form expression, or a recursive formula for this?

Thanks in advance!

$\endgroup$
2
2
$\begingroup$

Without loss of generality, the problem is equivalent to computing the probability $P(X_1 \geq \max(X_2, \ldots, X_n))$. We can transform the coordinates as $X_2-X_1, X_3-X_1, \ldots, X_n-X_1$ which is a fully general multivariate normal distribution of degree $n-1$. The problem is thus equivalent to finding the orthant probability for which there is no known closed-form solution for $n-1>3$

$\endgroup$
2
  • $\begingroup$ Arthur, thank you very much for your answer. Do you happen to have a reference for the last sentence? $\endgroup$ Dec 31 '13 at 7:29
  • $\begingroup$ Take your pick: google.com/… I don't think there is a proof that no closed-form solution exists, it's just that none is known and it seems very unlikely that one would exist. $\endgroup$
    – Arthur B
    Dec 31 '13 at 16:46

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.