Consider a $d$-dimensional Brownian motion $B=\left(B_1,...,B_d\right)$ whose components are independent and let $A$ be a $d\times d$ squared matrix such that $\sum_{i=1}^dA_{ii}^2=1$. Define the $W=A\,B$ the Brownian motion whose correlation matrix is $\rho=A\,A^{\dagger}$, where $\dagger$ indicate the transpose. Consider a $d$-dimensional process $\lambda_t\in\mathbb{L}_{loc}^2$ and a process $\eta_t$ that solves the stochastic differential equation

$$ d\eta_t=-\eta_t\,\lambda_t\cdot dW_t,\quad \eta_0=1, $$

where $\cdot$ indicates Euclidean product. So $\eta_t$ is a one-dimensional stochastic process. In the book by A. Pascucci it is written that the (unique) solution is

$$ \eta_t = \exp\left(-\int_0^t\lambda_s\cdot dW_s-\frac{1}{2}\,\int_0^t\left\langle\rho\,\lambda_s,\lambda_s\right\rangle\,ds\right). $$

I do not understand how exactly the co-variation $\left\langle\rho\,\lambda_s,\lambda_s\right\rangle$ is defined. It should, in principle, returns a scalar value since $\eta_t$ is scalar.

At some point in the book the co-variation between two $\mathbb{R}^d$-valued processes $X$ and $Y$ is defined as

$$ \left\langle X,Y\right\rangle_t=\left(\left\langle X^i,Y^j\right\rangle\right)_{i j=1,...,d} $$

so it seems that $\left\langle\rho\,\lambda_s,\lambda_s\right\rangle$ is a matrix. Does somebody have some clue?


This is only a matter of notation, by noting $cov_s=\left\langle\rho\,\lambda_s,\lambda_s\right\rangle$ this can be re-ritten written in matrix notation with the d-dimensional vector $\lambda_s$ and the $d\times d $ matrix $\rho$ in the following way which might seem more amenable to you :


which is a scalar.

Best regards

| cite | improve this answer | |

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.