2
$\begingroup$

suppose we have three variables here, x,y, z

now, what we know is that the correlation between x and z is 0.6, the correlation between y and z is 0.65. Here is the question, is there any formula to combine x and y , say combine(x,y), and there is stronger correlation between combine(x,y) and z

$\endgroup$
1
  • $\begingroup$ Hardly MO stuff, please try more adapted fora. $\endgroup$
    – Did
    May 8, 2013 at 11:13

1 Answer 1

0
$\begingroup$

Sometimes, it is possible to combine $X$ and $Y$ to get stronger correlation with $Z$. For example, if $Z=X+Y$, then $\mathrm{corr}(X,Y)$ and $\mathrm{corr}(X,Z)$ might be small, but $\mathrm{corr}(X+Y,Z)=1$. However, sometimes no combination $f(X,Y)$ will be better correlated with $Z$ than $Y$.

To see this, suppose $U$, $X$ and $Z$ are all random variables over $\{0,1\}$, and consider all functions $\{0,1\}\times\{0,1\}\rightarrow\{0,1\}$, of which there are $2^4$. Take the function $f$ such that $f(U,X)$ is maximally correlated with $Z$, and define $Y:=f(U,X)$. Since every combination of $X$ and $Y$ can be viewed as a combination of $U$ and $X$, we know that $Y$ is the combination that's most correlated with $Z$.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.