6
$\begingroup$

If I can generate normal random numbers in $N(0,1)$, how can I generate two dependent random numbers, $Y_1$ and $Y_2$ with means $\mu_1$, $\mu_2$ and $\sigma_1$, $\sigma_2$ and correlation coefficient $\rho$ respectively? Is there a known method for this problem?

$\endgroup$
8
$\begingroup$

If $X_1$ and $X_2$ are i.i.d. standard gaussian, then $X_3=\rho X_1+\sqrt{1-\rho^2}X_2$ is standard gaussian as well and the correlation of $X_1$ and $X_3$ is $\rho$. Hence, starting from $X_1$ and $X_2$, one can use $$ Y_1=\mu_1+\sigma_1X_1,\quad Y_2=\mu_2+\sigma_2\left(\rho X_1+\sqrt{1-\rho^2}X_2\right). $$

$\endgroup$
  • $\begingroup$ Thanks! Just a quick question, how can we prove that the correlation of $X_1$ and $X_3$ is $\rho$ then? $\endgroup$ – LBJ Feb 29 '12 at 21:11
  • $\begingroup$ Since $X_1$ and $X_3$ are centered and have unit variance, their correlation is $E(X_1X_3)$, which one computes by expanding the product. $\endgroup$ – Did Feb 29 '12 at 21:21
  • $\begingroup$ Got it again...I'm so stupid... $\endgroup$ – LBJ Feb 29 '12 at 21:22

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.