The Levy-Khintchine formula gives a triple $(a,\sigma,\nu)$ for the characteristic exponent $\Psi(s)$ of an infinitely divisible random variable where

$\Psi(s)=ias + \frac{1}{2}\sigma^2s^2 + \int_{\mathbb{R}}(1-e^{isx} + is\mathbb{1}_{|x|<1})d\nu(x)$

My question is whether $\nu$ is unique? Or whether it is only the unique Levy measure?

Are there references in the literature to this fact.



Yes, there is a one-to-one correspondance between infinitely divisble distributions and the triple $(a,\sigma,\nu)$ given by the Lévy-Khintchine formula. You can have a look at Lévy Processes and Infinitely Divisible Distributions by Ken Iti Sato (Theorem 8.1).

  • $\begingroup$ Thanks you. Is it possible that $\nu(x)$ is decreasing for certain values of x? Obviously this would imply that the random variable is not infinitely divisible. $\endgroup$ – Fantastic Mr. Fox Apr 24 '12 at 13:33
  • $\begingroup$ I don't understand what you're asking, since $\nu$ is a measure. What do you mean by decreasing? $\endgroup$ – Stefan Hansen Apr 24 '12 at 14:07
  • $\begingroup$ Sorry if $\nu(x)$ is negative for certain x? $\endgroup$ – Fantastic Mr. Fox Apr 24 '12 at 15:22
  • $\begingroup$ No, it's a measure, i.e. $\nu(A)\geq 0$ for every Borel set $A$. $\endgroup$ – Stefan Hansen Apr 25 '12 at 6:30
  • 1
    $\begingroup$ I don't follow what you are asking. In the Levy-Khintchine setup $\nu$ is a measure and $d\nu(x)$ denotes integration with respect to the measure $\nu$. Writing $\nu(x)$ does not make any sense in the setup, because $\nu$ is function on the Borel sets of $\mathbb{R}$. Nevertheless, I have no idea whether this generalizes to $\nu$ being a signed measure (i.e. $\nu$ is allowed to be negative). $\endgroup$ – Stefan Hansen Apr 30 '12 at 19:05

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.