Given a standard brownian motion $W_t$ and defining $\tau$ as:

$\tau :=\inf\{t\geq0:W_t=1$ or $W_t=-2\}$

The proof below shows that the stopping time is finite:

$$\begin{align*} P(\tau < t) &\geq P(|W_t|>2) \\ &= 1-P(|W_t| \leq 2)\\ &\geq1-4\frac{d}{dt}P(W_t \leq t)|_{t=0} \\ &=1-\frac{4}{\sqrt{2 \pi t}}\\ &\rightarrow 1 \qquad \text{as $t\rightarrow \infty$} \end{align*}$$

It's all staighforward except the line were the derivative is used:

$\geq1-4\frac{d}{dt}P(W_t \leq t)|_{t=0}$

How does this line relate to the line above?

  • $\begingroup$ It doesn't even make sense: if the derivative in that line is to be evaluated at $t=0$ then the value of that line should not depend on $t$, and it's nonsense to compare it to the preceding and succeeding lines which do. $\endgroup$ – Nate Eldredge Oct 20 '15 at 14:14
  • 1
    $\begingroup$ It also seems pointless: it's very easy to compute explicitly that $P(|W_t| \le 2) \to 0$. For instance, by scaling we have $P(|W_t| \le 2) = P(|W_1| \le 2/\sqrt{t})$. As $t \to \infty$, by "continuity from above" (a consequence of countable additivity), this converges to $P(|W_1| \le 0) = 0$. $\endgroup$ – Nate Eldredge Oct 20 '15 at 14:16
  • $\begingroup$ @Nate well it has me pretty confused too! $\endgroup$ – Bazman Oct 20 '15 at 14:18
  • 1
    $\begingroup$ Where did you find this "proof"? $\endgroup$ – Nate Eldredge Oct 20 '15 at 14:19
  • 2
    $\begingroup$ Well, assuming you've transcribed it correctly, I think it's at best a typo (though for what I don't know) and at worst it's just nonsense. If the book has more stuff like that I'd stop reading it. (Incidentally, I looked it up on Google Books to try to read the passage for myself, but you can't view any of the pages. But I did note that the book's subject is listed there as "Fiction / Romance / Historical / General" :-) $\endgroup$ – Nate Eldredge Oct 20 '15 at 14:28

As @NateEldredge has aleady pointed out, this particular line doesn't make sense and, moreover, the convergence

$$\mathbb{P}(|W_t| \leq 2) \xrightarrow[]{t \to \infty} 0$$

can be proved much easier using the scaling property.

However, here is a way to fix the proof: Obviously,

$$\mathbb{P}(|W_t| \leq 2) = \mathbb{P}(W_t \leq 2)-\mathbb{P}(W_t \leq -2) = \int_{-2}^2 \frac{d}{dx} \mathbb{P}(W_t \leq x) \, dx.$$

Using $W_t \sim N(0,t)$, we get

$$\mathbb{P}(|W_t| \leq 2) \leq \frac{4}{\sqrt{2\pi t}} \sup_{x \in [-2,2]} \exp(-x^2/2t) \leq \frac{4}{\sqrt{2\pi t}}.$$

This gives

$$\mathbb{P}(\tau<t) \geq 1- \mathbb{P}(|W_t| \leq 2) \geq 1- \frac{4}{\sqrt{2\pi t}}.$$


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.