If I have uncorrelated variables X,Y,Z how do I find the correlation between (X+Y) & (Y+Z)?

Ps. It's a theory question: ie, I can't compute the correlation from observed instances of X,Y & Z. What I can do is assume these variables have a standard deviation of a,b & c, and figure out the correlation of M & N (where M=X+Y & N=Y+Z) from there.

  • 2
    $\begingroup$ Coul you specify in your question what you mean by "and so on"? + maybe you could first simply compute the covariance between $X+Y$ and $Y+Z$ and see where it goes... $\endgroup$ – Vincent Guillemot Jun 16 '15 at 15:38
  • $\begingroup$ Edited the question. Essentially, I need to know how to compute covariance between (X+Y) & (Y+Z) given sd of X,Y & Z. $\endgroup$ – Retz Jun 16 '15 at 15:53
  • 3
    $\begingroup$ Hint: Covariance is a bilinear function meaning that \begin{align}\operatorname{cov}(aW+bX,cY+dZ) &= (ac)\cdot\operatorname{cov}(W,Y)+(bd)\cdot\operatorname{cov}(X,Z)\\& ~~+(ad)\cdot\operatorname{cov}(W,Z)+(bc)\cdot\operatorname{cov}(X,Y)\end{align} unless, of course, those $\&$'s mean something else. $\endgroup$ – Dilip Sarwate Jun 16 '15 at 16:00
  • $\begingroup$ Ok, I did not mean "compute" with a calculator, I meant: write down the definition of the correlation, or maybe first of the covariance, expand and simplify the terms that go away thanks to your hypotheses, and see what is left. $\endgroup$ – Vincent Guillemot Jun 16 '15 at 16:15
  • $\begingroup$ In other words: where are you stuck? $\endgroup$ – Vincent Guillemot Jun 16 '15 at 16:16

Use bilinearity and calculate $$ \DeclareMathOperator{\cov}{\mathbb{C}ov} \DeclareMathOperator{\var}{\mathbb{V}ar} \DeclareMathOperator{\cor}{corr} \cov(X+Y,Y+Z) =\cov(X,Y)+\cov(X,Z)+\cov(Y,Y)+\cov(Y,Z) =0+0+\var(Y)+0 $$ so $$ \cor(X+Y,Y+Z)=\frac{\cov(X+Y,Y+Z)}{\sqrt{\var(X+Y) \var(Y+Z)}} = \\ \frac{\var(Y)}{\sqrt{(\var(X)+\var(Y))(\var(Y)+\var(Z))}} $$ and you can take it from there.


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.