1
$\begingroup$

The fractional Brownian motion with Hurst Paramter $H\in(0,1)$ is a Gaussian Process $\{X(t),t\ge0\}$ with mean $0$ and covariance function $$\gamma(t,s)=1/2(|t|^{2h}+|s|^{2H}-|t-s|^{2H}).$$

I want to show that the fractional Brownian motion with Hurst parameter $H \in (0,1)$ is self-similar, that means for every fixed $c>0$ the processes $\{X(t),t\ge0\}$ and $\{c^{-H}X(ct),t\ge0\}$ have the same finite dimensional distributions.

I showed that $E[X(t)]=E[c^{-H}X(ct)]$ and $\operatorname{Cov}[X(t),X(s)]=\operatorname{Cov}[x^{-H}X(ct),c^{-H}X(cs)]$.

Does this mean they are equal in distribution? And therefore they have the same finite dimensional distributions which implies self similarity?

Or do I also have to show that $\{c^{-H}X(ct),t\ge0\}$ fulfills the $4$ properties of a Brownian motion? I mean $W(t)=0$, independent and stationary increments, Gaussian with mean $0$ and variance $t$, and continuous paths?

$\endgroup$
1
$\begingroup$

Recall that Gaussian distributions are uniquely characterized by their mean and covariance:

If $U:=(U_1,\ldots,U_n)$ and $V:=(V_1,\ldots,V_n)$ are Gaussian random vectors then $U=V$ in distribution if, and only if,

  • $\mathbb{E}(U_j)=\mathbb{E}(V_j)$ for all $j=1,\ldots,n$
  • $\text{cov}(U_i,U_j) = \text{cov}(V_i,V_j)$ for all $i,j=1,\ldots,n$.

Now let $0 \leq t_1 < \ldots < t_n$ be arbitrary times and set $$U := (X(t_1),\ldots,X(t_n)) \quad \text{and} \quad V := (c^{-H} X(ct_1),\ldots,c^{-H} X(ct_n))$$

for the fractional Brownian motion $(X(t))_{t \geq 0}$. Since $(X(t))_{t \geq 0}$ and $(c^{-H} X(ct))_{t \geq 0}$ are Gaussian processes, we know that $U$ and $V$ are Gaussian random vectors. Moreover, you have shown that the random vectors have identical mean vector and covariance matrix. Applying the above result we get $U=V$ in distribution. Since $0 \leq t_1 < \ldots < t_n$ are arbitrary, this shows that $(X(t))_{t \geq 0}$ and $(c^{-H} X(ct))_{t \geq 0}$ have the same finite-dimensional distributions.

$\endgroup$

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.