1
$\begingroup$

If there are independent random variables $X_1,\ldots,X_3, D(X_i)=\sigma^2, i=1,\ldots,3$

Then find correlation and covariance between variables $X_1+2X_2+X_3$ and $2X_1-X_2+6X_3$. I have no idea how to do it. But here is my take: $$\operatorname{Cov}(X_1+2X_2+X_3,2X_1-X_2+6X_3)$$

Here is my solution:

$$Cov(X_1,2X_1-X_2+6X_3)+Cov(2X_2,2X_1-X_2+6X_3)+Cov(X_3,2X_1-X_2+6X_3)=Cov(X1,2X_1)-Cov(X_1,X_2)+Cov(X_1,6X_3)+Cov(2X_2,2X_1)-Cov(2X_2,X_2)+Cov(2X_2,6X_3)+Cov(X_3,2X_1)-Cov(X_3,X_2)+Cov(X_3,6X_3)=2Var(X_1)-2Var(X_2)+6Var(X_3)=6\sigma^2$$ Can someone help me?

$\endgroup$
  • $\begingroup$ Hint: covariance is linear in both its arguments. Moreover it takes value is zero if the arguments are independent. $\endgroup$ – drhab Dec 29 '19 at 20:49
  • $\begingroup$ More specific: Covariance is bilinear. $\endgroup$ – callculus Dec 29 '19 at 20:57
  • 1
    $\begingroup$ @callculus Maybe some of you have another hint, I'm really not getting it. $\endgroup$ – user Dec 29 '19 at 21:31
  • $\begingroup$ $=2Cov(X_1,X_1)-Cov(X_1,X_2)+6Cov(X_1,X_3)+\cdots+6Cov(X_3,X_3)=...$ $\endgroup$ – Jean Marie Dec 29 '19 at 22:51
  • $\begingroup$ @user If have any further question, feel free to ask. $\endgroup$ – callculus Dec 29 '19 at 23:56
2
$\begingroup$

I show you the case with a linear combination of two variables. It a bit similar like multiplying out of ordinary brackets.

$$ \begin{align} & cov(a_1X_1+b_1X_2, a_2X_1+b_2X_2) \\\\ & \\\\ &=cov(a_1X_1, a_2X_1+b_2X_2)+cov(b_1X_2, a_2X_1+b_2X_2) \\\\ & \\\\ & =cov(a_1X_1, a_2X_1)+cov(a_1X_1,b_2X_2)+cov(b_1X_2, a_2X_1)+cov(b_1X_2, b_2X_2)\\\\ & \\\\ &=a_1a_2\cdot cov(X_1,X_1)+a_1b_2\cdot cov(X_1,X_2)+a_2b_1\cdot cov(X_1,X_2)+b_1b_2\cdot cov(X_2,X_2) \end{align} $$

Due the independence of $X_1$ and $X_2$ we have $cov(X_1,X_2)=0 $. And the covariance of $X_1$ and $X_1$ is the variance of $X_1$.

$$ \begin{align} & =a_1a_2\cdot Var(X_1)+b_1b_2\cdot var(X_2) \end{align} $$

You just have to transfer the calculation/result to the case of three random variables.


$$corr(Y_1,Y_2)=\frac{cov(Y_1,Y_2)}{\sqrt{var(Y_1)\cdot var(Y_2)}}$$

The term of the numerator is already calculated. And

$var(Y_i)=var(aX_1+bX_2+cX_3)=a^2var(X_1)+b^2var(X_2)+c^2var(X_3)$,

if the $X_i's$ are independent.

$\endgroup$
  • $\begingroup$ I added my result, wanted to check if I am correct. $\endgroup$ – user Jan 2 at 19:47
  • $\begingroup$ @user It looks perfect. $\endgroup$ – callculus Jan 2 at 20:01
  • $\begingroup$ #callculus Ok, and what about correlation, how to find that? $\endgroup$ – user Jan 2 at 20:18
  • $\begingroup$ @user I´ve made an edit. $\endgroup$ – callculus Jan 2 at 23:48

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.