2
$\begingroup$

Given that $\{B_t,t\ge0\}$ is a standard Brownian process. What is the conditional distribution of $B(s)$ given $B(t_1)=x_1$ and $B(t_2)=x_2$, for $0<t_1<s<t_2$?

My try: First i tried to write $B(s)=B(s)-B(t_1)+B(t_1)$ and then try to find the conditional distribution $\Pr(B(s),B(t_1)=x_1,B(t_2)=x_2)$ then we get $P(B(s)-x_1+x_1,B(t_1)=x_1,B(t_2)=x_2)$-but then i have no idea how to proceed. I am thinking of a 2nd apporach, using the moment generating function but seems that it is not easy to get the moment generating function of a condition random variable, as i don't even know how its distributed.

$\endgroup$
  • $\begingroup$ Would you happen to know the fact that $(B_s)_{t_1\leq s\leq t_2}\mid B_{t_1}=x_1,\,B_{t_2}=x_2$ is a Brownian bridge? $\endgroup$ – Stefan Hansen Dec 21 '12 at 10:29
  • $\begingroup$ @StefanHansen I heard about the brownian bridge but don't know that the above example is a Brownian bridge $\endgroup$ – Mathematics Dec 21 '12 at 16:56
1
$\begingroup$

One approach uses joint distributions to calculate the conditional distribution:

Since $(B_t)_{t \geq 0}$ is a Brownian motion we know that: $$\begin{equation*} (B_{t_1},B_s,B_{t_2}) \sim N(0,C) \qquad \text{mit} \, \, C:= \begin{pmatrix} t_1 & t_1 & t_1 \\ t_1 & s & s \\ t_1 & s & t_2 \end{pmatrix} \end{equation*}$$ The density of the joint distribution is given by$$\begin{equation*} p_{t_1,s,t_2}(x) = \frac{1}{(2\pi)^{\frac{3}{2}}} \cdot\frac{1}{\sqrt{(t_2-s) \cdot (s-t_1) \cdot t_1}} \cdot \exp \left(- \frac{1}{2}\cdot\left( \frac{(x_3-x_2)^2}{t_2-s} + \frac{(x_2-x_1)^2}{s-t_1}+ \frac{x_1^2}{t_1} \right) \right) \end{equation*}$$ where $x=(x_1,x_2,x_3)$. Now we can calculate the conditional density $$\begin{align*} p_{s|t_1,t_2}(x) &= \frac{p_{t_1,s,t_2}(x_1,x_2,x_3)}{p_{t_1,t_2}(x_1,x_3)} \\ &= \frac{1}{\sqrt{2\pi}} \cdot \sqrt{\frac{t_2-t_1}{(t_2-s) \cdot (s-t_1)}} \cdot \exp \left( - \frac{1}{2} \cdot \left( \frac{(x_3-x_2)^2}{t_2-s} + \frac{(x_2-x_1)^2}{s-t_1} - \frac{(x_3-x_1)^2}{t_2-t_1} \right) \right) \end{align*}$$ where $p_{t_1,t_2}$ denotes the joint distribution if $(B_{t_1},B_{t_2})$. Hence $$\mathbb{P}[B_s \in B|B_{t_1}=y_1,B_{t_2}=y_3] = \int 1_B(x) \cdot p_{s|t_1,t_2}(y_1,x,y_3) \, dx$$

$\endgroup$

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.