1
$\begingroup$

I have recently started to learn about Lévy processes, and have learnt about the Lévy-Khintchine theorem. My question is about the Poisson process, it stated to have Lévy triplet (0,0,$\lambda\delta(1)$), where $\delta(1)$ is the dirac measure with unity at 1, and zero elsewhere. But I can't seem to find an example showing this, is the Lévy triplet derived more from intuition about the process than through the Lévy-Khintchine theorem?

$\endgroup$
0
$\begingroup$

The Lévy - Kinchine theorem tells you that the characteristic function of a Lévy process has a certain form. Not how to compute this. For the Poisson process it is an easy exercise to compute the characteristic triplet:

$$\mathbb{E} e^{i \xi N_1} = \sum_{k \le 0} e^{i \xi k} \frac{ \lambda^k}{k!} e^{- \lambda} = exp( \lambda( e^{i \xi} - 1))$$

Taking the logarithm we get our representation:

$$\lambda( e^{i \xi} - 1) = \int \lbrace e^{i \xi u} - 1 \rbrace \ \lambda \delta_1(du)$$

Eventually I think it is important to understand that the Lévy - Kinchine theorem tells you that a general Lévy process actually behaves quite like a Poisson process, up to a drift and a Brownian motion. In fact the term involving the Lévy measure is always a compound Poisson Process unless the measure is infinite.

$\endgroup$
  • $\begingroup$ Are the triplets unique? Or can there be a characteristic be describe by the Lévy-Khintchine form with different triplet values? $\endgroup$ – df12 Feb 14 '17 at 13:45
  • $\begingroup$ Yes the triplets are unique. For example suppose that you have a triplet such that the associated $\psi$ is zero. Now the second order term prevails on the others at $\xi \to \infty$ so it has to vanish. Finally the term under the integral is not linear, thus both the remaining terms have to vanish as well. $\endgroup$ – Kore-N Feb 14 '17 at 16:20

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.