1
$\begingroup$

let's consider an sde of this kind:

\begin{equation} \label{eq:system} \begin{cases} dX_t=b(t,X_t,Y_t)dt+\sigma(t,X_t,Y_t)dW_t \\ X_0=x_0 \\ dY_t=B(t,X_t,Y_t)dt+C(t,X_t,Y_t)dW_t^2 \\ Y_0=y_0 \end{cases} \end{equation} where t $\in [0,T]$, $T$ is the time horizon , $x_0 \in \mathbb{R}^d$, $y_0 \in \mathbb{R}^l$, b, $\sigma$, B, C are Borel functions and $(\Omega, \mathcal{F}, \mathcal{F}_t,W_t , \mathbb{P})_{t \geq 0}$ is a continuous standard $\mathbb{R}^d$-Brownian motion and $\{ W_t^2 \}_{t \geq 0}$ is a continuous standard $\mathbb{R}^l$-Brownian motion on the same filtration having instantaneous correlation $\rho \in (-1,1)$ \begin{align*} \mathbb{E}[dW_t^i dW_t^{2,j}]=\rho dt && \forall i \leq d, j \leq l \end{align*} Do the classic theorem of existence and uniqueness of strong solution holds also for this system? Do you have any reference? Thank you all in advance!

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ When you say "correlated", do you mean they are still jointly Gaussian, i.e. they form a two-dimensional Brownian motion with covariance $\left(\begin{smallmatrix} 1 & \rho \\ \rho & 1 \end{smallmatrix} \right)$? If yes then by a linear transformation you can rewrite the system as driven by a standard 2-D BM and everything is fine. If no then things get much harder and I don't know what can be said. $\endgroup$
    – Nate Eldredge
    Commented May 7, 2020 at 15:26
  • $\begingroup$ yes, i mean jointly gaussian. But the BM are not 1D but d-dimensional and l-dimensional respectively. Does it work also in this sense? $\endgroup$
    – defex95
    Commented May 7, 2020 at 15:40
  • 1
    $\begingroup$ I missed that. In general it will work whenever they are a linear transformation of a standard BM, which would be true for any covariance between them. But are you sure the condition you state is possible? If I compute correctly, when $d=l=3$ and $\rho = 1/2$ the resulting covariance matrix is not positive definite. $\endgroup$
    – Nate Eldredge
    Commented May 7, 2020 at 15:51

1 Answer 1

Reset to default
1
$\begingroup$

As mentioned in the comments you simply need to do a transformation. Here in "Modelling the Stochastic Correlation" pg.23 they give the concrete transformation

enter image description here

You replace your SDE by the above. Then as usual you need to have regularity for the coefficients (see What work has been done on SDE with diffusion coefficients of bounded variation in $\mathbb R^d$? for references).

One reference I found online is section 3.3 in

"American-Type Options: Stochastic Approximation Methods, ".

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .