2
$\begingroup$

Assume that the random vector $(X,Y)$ is (bivariate) normally distributed. Show that $$ \Bbb E[X|Y=y]= \Bbb E[X]+ \frac {Cov[X,Y]}{Var[Y]}(y-\Bbb E[Y])$$

Also, $$ Var[X|Y=y]= (1-\rho^2) Var[X]$$

I know i should be converting these variables into standard normal and then using Cholesky decomposition to come up with independent standard normal, I am getting pretty close to the answer but, its not neat. I might have done something wrong, Can some one please lay out the first step to convert X&Y to standard normal?? Thanks so much

$\endgroup$
4
$\begingroup$

By Cholesky decomposition, you can express the normal random variables $X$ and $Y$ in the form \begin{align*} Y &= E(Y) + \sqrt{Var(Y)}\, \xi,\\ X &= E(X) + \sqrt{Var(X)}\left(\rho \xi+\sqrt{1-\rho^2} \eta\right), \end{align*} where $\rho = \frac{Cov(X, Y)}{\sqrt{Var(X)Var(Y)}}$ is the correlation, $\xi$ and $\eta$ are two independent standard normal random variables.

Then, \begin{align*} E(X \mid Y) &= E\left(E(X) + \sqrt{Var(X)}\left(\rho \xi+\sqrt{1-\rho^2} \eta\right) \mid \xi \right)\\ &=E(X) + \rho \sqrt{Var(X)}\xi\\ &=E(X) + \frac{Cov(X, Y)}{\sqrt{Var(Y)}}\xi\\ &=E(X) + \frac{Cov(X, Y)}{Var(Y)}\big(Y-E(Y) \big). \end{align*} The computation for $Var(X\mid Y)$ is similar, specifically, \begin{align*} Var(X \mid Y) &=E\left((X-E(X\mid Y))^2\mid Y \right)\\ &=E\left( (X-E(X\mid Y))^2\mid \xi\right)\\ &=E\left(Var(X)(1-\rho^2) \eta^2 \mid \xi\right)\\ &=E\left(Var(X)(1-\rho^2) \eta^2\right)\\ &=Var(X)(1-\rho^2). \end{align*}

$\endgroup$
3
$\begingroup$

Another approach

$$f_{X|Y}(x,y)=\frac{f_{X,Y}(x,y)}{f_{Y}(y)}\tag 1$$ Set $$u=\frac{x-\mu_X}{\sigma_X}$$ and $$v=\frac{y-\mu_Y}{\sigma_Y}$$ we have $$f_{X|Y}(x,y)=\frac{\frac{1}{2\pi\sigma_X\sigma_Y\sqrt{1-\rho^2}}\exp\left(-\frac{u^2-2\rho uv+v^2}{2(1-\rho^2)}\right)}{\frac{1}{\sqrt{2\pi}\sigma_Y}\exp\left(-\frac{1}{2}v^2\right)}\\\qquad\qquad\qquad\qquad=\frac{1}{\sqrt{2\pi(1-\rho^2)}\sigma_X}\exp\left(-\frac 12\left[\frac{u-\rho v}{\sqrt{1-\rho^2}}\right]^2\right)\tag 2$$ Indeed $$f_{X|Y}(x,y)=\frac{1}{\sqrt{2\pi(1-\rho^2)}\sigma_X}\exp\left(-\frac{1}{2}\left[\frac{x-(\mu_X+\rho\frac{\sigma_X}{\sigma_Y}(y-\mu_Y)}{\sigma_X\sqrt{1-\rho^2}}\right]^2\right)\tag 3$$ as a result $$\mathbb{E}[X|Y]=\mu_X+\rho\frac{\sigma_X}{\sigma_Y}(y-\mu_Y)\tag 4$$ and $$\text{Var}(X|Y)=\sigma_X^2(1-\rho^2)\tag 5$$

$\endgroup$
  • 1
    $\begingroup$ A nice elementary approach. $\endgroup$ – Gordon Dec 15 '16 at 20:01
  • $\begingroup$ So thanks Gordon . Indeed your answer is so nice $\endgroup$ – user16651 Dec 15 '16 at 20:13

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.