I am confused on the multi-dimensional Brownian motion.

$B_t$ is a standard Brownian motion based on a filtered probability space $(\Omega, \mathcal{F}, (\mathcal{F}_t)_{t \geq 0}, \mathbb{P})$ if $$Cov(B_t,B_s)=\mathbb{E}\left[B_tB_s\right]= \min (s,t)$$

When $B_t=(B_t^1, \cdots , B_t^d)$ is taking its values in $\mathbb{R}^d$,

the above equality is valid for each $B_t^i, \, i=1, \cdots , d$; Which is the same for all $B_t^i$ because $Cov(B_t^i,B_s^i)= \min (s,t)$ does not depend on $i$.

How can we write the variance-covariance matrix $C_{i,j}=Cov(B_t^i,B_s^j)$ in terme of $\min (s,t)$? $Diag(C_{i,j})= \min (s,t)$. What about the other termes?


In a multi-dimensional Brownian motion, the components are independent; this is normally part of the definition. So $\{B_t^i, t \ge 0\}$ is independent of $\{B_t^j, t \ge 0\}$ whenever $i \ne j$. Thus $Cov(B_t^i, B_s^j) = 0$ whenever $i \ne j$.

  • 2
    $\begingroup$ And $\mathrm{Cov}(B^i_t,B^j_s)=0$ (to answer the question asked by the OP). $\endgroup$ – Did Dec 18 '11 at 20:19
  • $\begingroup$ What about a fractional Brownian motion defined by it's covariance $$ R_{H}(t,s) := \frac{H}{2} ( {|t|}^{2{H}} + {|s|}^{2{H}} - {|t-s|}^{2{H}}) $$, there is no independance. $\endgroup$ – Zbigniew Dec 18 '11 at 20:39
  • 1
    $\begingroup$ $E[B_tB_s]$ means the expectation of the kronecker product between $B_t$ and $B_s$, so it is a matrix. If $B^i$ and $B^j$ are independent then the matrix contains a lot of zeros, if not, then it does not :) $\endgroup$ – Martingalo Mar 13 '14 at 7:57

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.