1
$\begingroup$

I am supposed to show this inequality for a 2 or 3 dimensional Brownian motion:

$\textbf{P}^0\{\sup\limits_{t\leq k}{|B(t)|}\geq\frac{1}{2}\}\leq 2\textbf{P}^0\{{|B(k)|\geq\frac{1}{2}}\}$ (where $\textbf{P}^0$ means the BM is started in zero).

This looks a lot like Doob's maximum inequality to me, but on the RHS there is a probabilty instead of an expectation and also $|B|$ is not a martingale for d=2,3. I think you can fix the last part by just taking $\ln|B|$ and $|B|^{-1}$ instead, but I don't see how to replace the expectation with a probability.

Thanks in advance!

$\endgroup$
  • 1
    $\begingroup$ en.wikipedia.org/wiki/Reflection_principle_(Wiener_process) $\endgroup$ – d.k.o. Jan 23 at 2:14
  • $\begingroup$ Thanks! But I thought the reflection principle was only applicable to linear Brownian motion? Can I also use it for higher dimensions? $\endgroup$ – John Doe Jan 23 at 16:55
  • $\begingroup$ What is $|B(t)|$ for higher dimensions? $\endgroup$ – d.k.o. Jan 23 at 23:06
  • $\begingroup$ I am not sure what you mean ... it's of course length of the random vector with independent Brownian motions in it's entries. But I don't think $|B(t)|$ is a Brownian motion again, is it ...? $\endgroup$ – John Doe Jan 24 at 0:08
  • $\begingroup$ Is it $\ell_2$ norm? $\endgroup$ – d.k.o. Jan 24 at 7:52
0
$\begingroup$

For $a>0$ let $\tau_a:=\inf\{t\ge 0:|B_t|=a\}$. Then $$ \mathsf{P}^0(\tau_a\le t)=\mathsf{P}^0(\tau_a\le t,|B_t|\ge a)+\mathsf{P}^0(\tau_a\le t,|B_t|< a).\tag{1} $$

Using the strong Markov property (e.g. Theorem 8.3.7 on page 314 here), \begin{align} \mathsf{P}^0(\tau_a\le t,|B_t|< a)&=\mathsf{E}^0[1\{\tau_a\le t\}\mathsf{P}^0(|B_{\tau_a+(t-\tau_a)}|<a\mid \mathcal{F}_{\tau_a})] \\ &=\mathsf{E}^0[1\{\tau_a\le t\}\mathsf{P}^{B_{\tau_a}}(|B_{t-\tau_a}|<a)]\le \frac{1}{2}\mathsf{P}^0(\tau_a\le t).\tag{2} \end{align}

Combining $(1)$ and $(2)$, we get $$ \mathsf{P}^0\left(\sup_{s\le t}|B_t|\ge a\right)=\mathsf{P}^0(\tau_a\le t)\le 2\mathsf{P}^0(|B_t|\ge a). $$

It remains to verify the last inequality in (2), i.e. for any $x$ with $|x|=a$ and $s\ge 0$, $$ \mathsf{P}^x(|B_s|<a)\le 1/2. $$

$\endgroup$

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.