I am given $Z_1$ and $Z_2$ independent standard normal variables, I have to find two random variables $X_1$ and $X_2$ with correlation $\operatorname{corr}(X_1,X_2) = p$, where $p\in (−1, 1)$.

Any help as to how I can get to the answer or which property or formula I should look into to get to this answer..


  • $\begingroup$ The formula you were given is $X_2=pZ_1+\sqrt{1-p^2}Z_2$, please correct your post. $\endgroup$ – Did Mar 13 '12 at 6:09
  • $\begingroup$ Still incorrect. $\endgroup$ – Did Mar 13 '12 at 6:36

We use the amended suggestion $X_1=Z_1$, $X_2=p Z_1+\sqrt{1-p^2}Z_2$.
The correlation coefficient of $X_1$ and $X_2$ is $$\frac{E(X_1X_2)}{\sqrt{\text{Var}(X_1)\text{Var}(X_2)}}.\qquad\qquad(\ast)$$ We first calculate the numerator of $(\ast)$. We have $$E(X_1X_2)=E\left(Z_1\left(p Z_1+\sqrt{1-p^2}Z_2\right)\right)=p E(Z_1^2) +\sqrt{1-p^2}E(Z_1Z_2).$$ Since $Z_1$ is standard normal, it has mean $0$ and variance $1$, so $E(pZ_1^2)=p$. By independence, $E(Z_1Z_2)=E(Z_1)E(Z_2)=0$. So $E(X_1X_2)=p$.

We now deal with the variances in the denominator of $(\ast)$. The variance of $X_1$ is $1$. For the variance of $X_2$, use the fact that $X_2=p Z_1+\sqrt{1-p^2}Z_2$, a linear combination of independent normals with variance $1$. So $\text{Var}(X_2)=p^2+(\sqrt{1-p^2})^2=1$.

It follows that the correlation coefficient of $X_1$ and $X_2$ is equal to $p$.

  • $\begingroup$ Hello Andre, In this question we dont have to prove $\operatorname{corr}(X_1,X_2) = p$ is equal to $X_1=Z_1$ and $X_2=p\cdot Z_1+ \sqrt{1-p^2}\cdot Z_2$. We have to find 2 random variables X1 and X2. I have editied the question for further clarity...any hint will be appreciated... Thanks $\endgroup$ – newbietolinalg Mar 14 '12 at 5:45
  • 1
    $\begingroup$ The answer I gave proved that if $X_1=Z_1$, $X_2=pZ_1+\sqrt{1-p^2} Z_2$, then the correlation coefficient turns out to be $p$. The proof used the basic formula $(\ast)$ for the correlation coefficient, plus some useful facts about the normal, in particular the formula for the variance of a linear combination of independent normals. There are other ways to produce $X_1$ and $X_2$ that have correlation coefficient $p$, but this one is probably the simplest. $\endgroup$ – André Nicolas Mar 14 '12 at 5:52

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.