Suppose $X_t$ solves the SDE $$ dX_t = \mu(t,X_t)dt +\sigma(t,X_t)dZ_t, $$ where $Z_t$ is a Lévy process on $\mathbb{R}^d$, $g(t,s,x):[0,T]\times[0,1]\times \mathbb{R}^d \rightarrow \mathbb{R}^d$ is $C^{1,2,2}$ and $S_t$ is an SDE solving $$ dS_t = a(t,S_t,X_t)dt + b(t,S_t,X_t)dY_t, $$ where $Y_t$ is a Lévy process as well.


Is it possible to obtain an SDE for: $$ \mathbb{E}[g(t,S_t,X_t)|\sigma(X_s)_{0\leq s\leq t}]? $$ That is the projection of $g(t,S_t,X_t)$ onto the information only known throught $X_t$ up to time $t$?


I was considering some type of filtering approach, but I'm not certain how to go about this.


You can obtain an SPDE (or, rather, an SIDE - stochastic integro-differential equation) for the conditional density of $S$ given the history of $X$. From here you can calculate the quantity you're looking for easily.

In the continuous case (i.e. $Z$ and $Y$ are Wiener processes), there are plenty of references, look for Zakai's equation. When you have jumps, the main reference I know of is B. Grigelionis and R. Mikulevicius: Nonlinear filtering equations for stochastic processes with jumps.

  • 1
    $\begingroup$ Fantastic!! Thank you so much for the references m7e! :D $\endgroup$ Nov 16 '16 at 18:37

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.