0
$\begingroup$

I'm studying about the reflection principle of the brownian motion, and I found that this result is a direct consequence of this principle:

Let $B_t$ a brownian motion, then for every $a \in \mathbb{R} \ $,

$$\mathbb{P}(\lim_{t \to \infty} \sup_{s\in [0,t]} B_s > a) = 1$$

I'm trying to prove this statement using the reflection principle but I'm totally lost. I can't see how are those results related.

|cite|improve this question
$\endgroup$
  • $\begingroup$ Which formulation of the reflection principle do you know/use? It is direct consequence of the reflection principle that $$M_t := \sup_{s \leq t} B_s$$ equals in distribution $|B_t$|. Knowing this, it shouldn't be difficult to prove the assertion. $\endgroup$ – saz Dec 7 '18 at 7:51
1
$\begingroup$

Let $M_t = \sup_{s \in [0,t]} B_s$. For $a < 0$ the statement is trivial, so take $a \ge 0$.

The reflection principle says that $P(M_t > a) = 2 P(B_t > a)$. Now since you know $B_t \sim N(0,t)$, you can compute $\lim_{t \to \infty} 2 P(B_t > a)$ and determine that it equals 1. Thus $\lim_{t \to \infty} P(M_t > a) = 1$.

This is not the same as the desired statement, but can be used to prove it in this case. Note that $M_t$ is increasing in $t$, so necessarily $\lim_{t \to \infty} M_t$ exists (as a random variable) and $\lim_{t \to \infty} M_t \ge M_u$ for any fixed $u$. Hence $P(\lim_{t \to \infty} M_t > a) \ge P(M_u > a)$. Passing to the limit as $u \to \infty$ and using what we previously proved, we have $P(\lim_{t \to \infty} M_t > a) \ge \lim_{u \to \infty} P(M_u > a) = 1$.

|cite|improve this answer
$\endgroup$
  • $\begingroup$ What is $\lim_{t\to\infty} M_t$? $\endgroup$ – thomasb Sep 20 at 16:37
  • $\begingroup$ It's the usual thing. It's the random variable $M$ defined as $M(\omega) = \lim_{t \to \infty} M_t(\omega)$. We know by definition of $M_t$ that for each $\omega$, $M_t(\omega)$ is an increasing function of $t$, so the limit exists in $[0,+\infty]$ for each $\omega$ and $M$ is well defined. It is an exercise to verify that it is measurable. Of course, we end up proving that $P(M > a) = 1$ for every $a$, from which it follows that in fact $M = +\infty$ almost surely. $\endgroup$ – Nate Eldredge Sep 20 at 17:45
0
$\begingroup$

From the Law of Iterated Logarithm: $$ \limsup_{t\to+\infty}\frac{B_t}{\sqrt{2t\ln\ln t}} \overset{\mathbb{P}\rm -a.s.}{=} 1, $$ so $$ \mathbb{P}\left(\lim_{t\to+\infty}\sup_{s\in[0,t]}B_s>a\right)=1. $$

|cite|improve this answer
$\endgroup$

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.