Consider a random variable $Y$ with $\mathbb{E}(Y) = \mu_Y$ and $\text{Var}(Y) = \sigma_Y^2$. Let $X$ be another random variable with $\mathbb{E}(X|Y) = A+BY$ and $\text{Var}(X|Y) = \sigma_{X|Y}^2$, where $A, B, \sigma_{X|Y}^2$ constant.

Compute $\rho = \frac{\text{Cov(X,Y)}}{\sigma_X\sigma_Y} = \frac{\mathbb{E}(XY) - \mathbb{E}(X)\mathbb{E}(Y)}{\sigma_X\sigma_Y}$.

My thoughts

I first calculated $\mathbb{E}(X)$ using conditional expectation, and found that $\mathbb{E}(X) = A+B\mu_Y$, thus $E(X|Y) = \mu_X + B(Y-\mu_Y)$. I then computed $\text{Var}(X) = \sigma_X^2$ using the conditional variance formula and found that $\text{Var}(X) = B^2\sigma_Y^2+\sigma_{X|Y}^2 = \sigma_X^2$, thus we can write $\sigma^2_{X|Y} = \sigma_X^2-B^2\sigma_Y^2$.

Therefore, I have $\mathbb{E}(X)$ and $\mathbb{E}(Y)$, but now I need to find $\mathbb{E}(XY)$ to calculate $\rho$. Since this is a follow up exercise, I assume that I need to use the conditional variance $\sigma_{X|Y}^2$, however I don't know how to proceed. Any thoughts?

  • $\begingroup$ $E(XY)=E(E(XY\mid Y))=E(Y\,E(X\mid Y))$. $\endgroup$ – StubbornAtom Jun 22 '18 at 9:58

By definition $\mathbb{E}X1_{B}\left(Y\right)=\mathbb{E}\left[\mathbb{E}\left[X\mid Y\right]1_{B}\left(Y\right)\right]$ for every Borel set $B$.

This can be extended to $\mathbb{E}Xf\left(Y\right)=\mathbb{E}\left[\mathbb{E}\left[X\mid Y\right]f\left(Y\right)\right]$ for suitable functions $f$.

Taking for $f$ the identity we find: $$\mathbb{E}XY=\mathbb{E}\left[\mathbb{E}\left[X\mid Y\right]Y\right]=\mathbb{E}\left[\left(A+BY\right)Y\right]=A\mu_{Y}+B\mathbb{E}Y^{2}$$

This leads to: $$\mathsf{Cov}\left(X,Y\right)=\sigma_{Y}^{2}$$


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.