5
$\begingroup$

This question might be a bit basic, but I am struggling to understand the connection between various versions of the Ito's lemma for Levy processes (and semimartingales in general). Could someone clarify what is the relationship between the jump measure, levy measure and if it is possible to express these quantities as a $$\sum_{\Delta X_t \neq 0} sth$$ in the context of Ito's lemma?

In particular I am looking at the application to the exponential function at the moment, if it simplifies anything.

EDIT: I am trying to link together the martingale representation theorem as in the answer to this question (Cont Tankov use the analogous notation): Martingale representation theorem for Levy processes

Under some integrability conditions, there exist $\phi$ and $\psi$ st:

$$M_t=M_0+\int^t_0 \phi(s)dW_s+\int^t_0 \int_{\textbf{R}}\psi(s,x)\tilde{N}(ds,dx)$$ Where $\tilde{N}(ds,dx)$ is the compensated measure of the Lévy process $X$.

and the Ito's formula for Levy processes (taken from Cont Tankov) is:

$$f(X_t) - f(X_0) = \int_0^t \frac{\sigma^2}{2} f''(X_s) ds + \int_0^t f'(X_{s-}) dX_s$$ $$ + \sum_{0\leq s \leq t} \textbf{1}_{(\Delta X_s \neq 0)}[f(X_{s}) - f(X_{s-}) - \Delta X_s f'(X_{s-})]$$

Please note that the formulas above are for scalar processes.

My question is how does the term $\int^t_0\int_{\textbf{R}}\psi(s,x)\tilde{N}(ds,dx)$ relate to $\sum_{0\leq s \leq t} \textbf{1}_{(\Delta X_s \neq 0)}[f(X_{s}) - f(X_{s-}) - \Delta X_s f'(X_{s-})]$ and to the Levy measure of the process $X$.

Also Proposition 8.16 from Cont-Tankov suggests that a funtion of a Levy process is a martinale iff we have:

$$f(X_t) - f(X_0) = \int_0^t f'(X_s)\sigma dWs + \int_{[0,t]\times R} \tilde{N}(ds,dx) [f(X_{s-} + y) - f(X_{s-})]$$

And again I am not sure how to relate the formula above to the Ito's formula.

$\endgroup$
4
  • $\begingroup$ @ Grzenio : Hi what is "sth" exactly representing in your sum ? $\endgroup$
    – The Bridge
    Aug 11, 2011 at 22:42
  • $\begingroup$ @Bridge, that is something I got confused about. I saw some versions of Ito's formula that contains the sum over all jumps of some expression involving the function, process and jump size. At the same time I saw a martingale representation theorem which is written using the jump measure. And my Levy process is defined as the triplet, so I have the Levy measure. I am trying to link these three, so that I can figure out when some function of a levy process is a martingale. $\endgroup$
    – Grzenio
    Aug 12, 2011 at 8:06
  • $\begingroup$ @Grzenio : If all the versions of the Itô's Lemma you have at your disposal are correct then all those formulations must be equivalent. I think you should edit your post with those versions written explicitly to clarify the issue. Regards $\endgroup$
    – The Bridge
    Aug 12, 2011 at 13:13
  • $\begingroup$ @Bridge, I tried to provide all the details in the text of the question. Thanks for looking. $\endgroup$
    – Grzenio
    Aug 13, 2011 at 15:38

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.