I need to prove the following:

Let $X$ be a Brownian motion with drift $\mu$ and volatility $\sigma$. Pick three time points $s < u < t$. Then, the conditional distribution of $X_u$ given $X_s = x$ and $X_t = y$ is normal; in fact $$(X_u\mid X_s = x, X_t = y)\sim \mathcal{N}\left(\frac{t-u}{t-s}x+\frac{u-s}{t-s}y,\sigma^2\frac{(t-u)(u-s)}{t-s}\right)$$

I know the conditional distributions of Gaussian are also Gaussian. I also have formulae to find $\mathbb E(X\mid Y=y)$ and $\operatorname{Var}(X\mid Y=y)$ when $X$ and $Y$ are jointly Gaussian. However I cannot find a way to apply those results here. Could someone help me please?


Consider this theorem on conditional expectation of multivariate Gaussian distribution. Let $x=X_u$ and $Y=\left[X_s \quad X_t\right]^\top$. Let $X_0=0$, so that $\mu_X=0$ and $\mu_Y=\left[0 \quad 0\right]^\top$.

\begin{align} \Sigma_{XX}&=Cov(X_u,X_u)=\sigma^2u\\ \Sigma_{XY}&=E(XY^\top)=[E(X_uX_s)\quad E(X_uX_t)]=\sigma^2[s\quad u]\\ \Sigma_{YY}&=E(YY^\top)=\left[\begin{matrix}E(X_s^2) & E(X_sX_t) \\ E(X_tX_s) & E(X_t^2)\end{matrix}\right]=\sigma^2\left[\begin{matrix}s & s \\ s & t\end{matrix}\right] \end{align}

Plugging the formula, the random variable $X_u\mid (X_s=x,X_t=y)$ is also Gaussian with

  • mean: $\Sigma_{XY}\Sigma_{YY}^{-1}\left[\begin{matrix}x \\ y\end{matrix}\right]=\frac{t-u}{t-s}x+\frac{u-s}{t-s}y$
  • variance:$\Sigma_{XX}+\Sigma_{XY}\Sigma_{YY}^{-1}\Sigma_{YX}=\sigma^2u+\sigma^2[s\quad u]\left[\begin{matrix}s & s \\ s & t\end{matrix}\right]^{-1}\left[\begin{matrix}s \\ u \end{matrix}\right]=\sigma^2\frac{(t-u)(u-s)}{t-s}$

You need to apply this theorem about how on conditioning a Gaussian vectors.

You are in the situation where you have this 3-d Gaussian vector $V=(X_s,X_u,X_t)$.

You will have to compute the mean vector and covariance matrix of $V$, and then applying the theorem to your situation.



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.