Conditional expectations and joint distributions

37 Views Asked by At

Let $Z, W, U$, be standard independent Gaussian and three moments in time $u < s < t$. Let us define:

$$B_u = \mu u + σ \sqrt uZ$$ $$B_s = B_u + (s − u)\mu + σ \sqrt{ (s-u)}W$$ $$B_t = B_s + (t − s)\mu + σ\sqrt{ (t − s)}U$$

I have to compute:

1) $E[B_sB_t\mid B_t]$.

2) $E[B_s^2B_u\mid B_t]$.

3) The joint distribution of $B_s$ and $B_t$.

Here is my attempt:

1) Knowing $B_t$ implies knowing $B_s$ so that: $E[B_sB_t\mid Bt]=B_sB_t$.

2) Knowing $B_t$ implies knowing $B_s$ and $B_u$ so that: $E[B_s^2B_u\mid Bt]=B_s^2B_u$.

3) Because $Z, W, U$ are independent, I know that $B_u, B_s, B_t$ are normal, but $B_s$ and $B_t$ are not independent gaussian so I do not have any idea about how their joint distribution looks like.

Any help would be appreciated.

1

There are 1 best solutions below

2
On

For 1) and 2) consider the answer from TheSimpliFire except the fact that $$E[U|B_t] = 0$$ what's not true in general.

But you should be able to get $E[U|B_t]$ by considering:

$$\sqrt{t-s}\cdot E[U|B_t] + \sqrt{s-u} \cdot E[W|B_t] + \sqrt{u} \cdot E[Z|B_t] = \sqrt{t-s}\cdot U + \sqrt{s-u} \cdot W + \sqrt{u} \cdot Z$$

and think about a relation between all three conditional expectations.

For 3) consider that $B_s$ is Gaussian and $$X = (t − s)\mu + σ\sqrt{ (t − s)}U$$ as well and $X$ is independent from $B_s$. Additionally the joint distributin of $(B_s,B_t)$ is the distribution from $(B_s,B_s + X)$ and the second one can be calculated using the independence of $B_s$ and $X$ via:

$$\begin{align*}P(B_s \le a, B_t \le b) &= P(B_s \le a, B_s + X \le b) \\ &= E[1_{\{B_s < a\}}1_{\{B_s + X < b\}}] \\ &= E[f(X,B_s)]\end{align*}$$ with $$f(x,y) = 1_{\{y < a\}}1_{\{y+x < b\}}$$

And this expectation can be calculated using the joint density of $B_s$ and $X$ which is easy because $B_s$ and $X$ are Gaussian and independent.