2
$\begingroup$

This problem cropped up in some research I am doing. I imagine it is standard, but I cannot seem to find the answer.

Let $W_t$ be a standard Brownian motion. Suppose there are four values $a < 0 < b$ and $c<d$. For a given $t > 0$, I want to know the probability that $W_t$ reaches $b$ before reaching $a$ and then after reaching $b$ never falls below $c$ nor exceeds $d$ until time $t$.

Thanks all.

$\endgroup$
  • $\begingroup$ If it's a standard Brownian motion, then the probabilty that after some point it never falls below $c$ is $0$. $\endgroup$ – Michael Hardy Aug 12 '13 at 22:46
  • $\begingroup$ Sorry, I want to know the prob. it never falls below $c$ or exceeds $d$ until time $t$. $\endgroup$ – Brian Aug 12 '13 at 22:51
2
$\begingroup$

I can't add comments (not enough reputation) so I'm posting this as an answer :

Consider your first problem which is reaching $b$ before $a$. Denote $T_a$ and $T_b$ stopping times of $W_t$ reaching a and b and $T=\min(T_a,T_b)$. $W_t$ being a martingale (stopping theorems look wikipedia): $\mathbb{E}[W_{\min(t,T)}]=0=\mathbb{E}[a1_{T_a<T_b}]+\mathbb{E}[b1_{T_b\leq T_a}]$ And then : $$\mathbb{P}[T_a<T_b]=\frac{b}{b-a}$$

You can follow this method to get your results

$\endgroup$
  • $\begingroup$ That's not quite what I need. In addition to the probability that $T_b < T_a$, I also need to know the probability that $W_t$ remains within the interval $[c,d]$ (where $c < b < d$) during $[T_b,t]$. $\endgroup$ – Brian Aug 13 '13 at 0:38
  • $\begingroup$ Agreed, but the result you are looking for can be found in a similar manner. My answer is just the very basic case. What have you tried so far ? $\endgroup$ – Bertrand R Aug 13 '13 at 1:16
  • $\begingroup$ I can compute $P[T_b < T_a]$, $P[T_b < t]$, and $P[\{T_c > t\}\cap \{T_d > t\}]$ but since these events are not independent, I cannot seem to find their joint probability. $\endgroup$ – Brian Aug 13 '13 at 2:30
  • $\begingroup$ Hint : $W_{t+T_b}-W_{T_b}$ is a standard brownian motion and then reduce to the previous case. (edit : of course you need $c<b<d$ otherwise the result is trivial) $\endgroup$ – Bertrand R Aug 13 '13 at 2:32
  • $\begingroup$ Sorry for being slow, but I still don't see it. Define $f(t) = P(T_b \in dt,T_b<T_a)$ and $G_x(s) = P[\min_{0\leq j \leq s} W_j \geq c, \max_{0\leq j \leq s} W_s \leq d | W_0 = x]$. Then the way I would compute what I'm seeking if I knew $f(t)$ is $\int_0^t f(s)G_b(t-s)ds$. $\endgroup$ – Brian Aug 13 '13 at 2:48

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.