Given $X_1$ and $X_2$ as correlated ($\rho$) and jointly Normal random variables (each distributed ~$N(0,\sigma_X^2)$), and also given $Y$ as Normally distributed((~$N(0,\sigma_Y^2)$)) and independent of $X_1$ and $X_2$, how can one compute

$p(X_2-X_1+Y>\pi, X_2>\pi)$

Any help and guidance would be greatly appreciated!

I know this is equivalent to

$p(X_2-X_1+Y>\pi, X_2>\pi)$=$p(X_2-X_1+Y>\pi | X_2>\pi)p(X_2>\pi)$

but I'm not sure if this is the way to proceed and how to do so. Perhaps there is a simple way to do this that I'm missing? I would really appreciate the help!


This is $$ E\left(\Phi\left(\frac{X_2-X_1-\pi}{\sigma_Y}\right)\cdot\mathbf 1_{X_2\gt\pi}\right). $$ Now, use the density of $(X_1,X_2)$ to transform this expectation into a double integral on $\mathbb R\times(\pi,+\infty)$.

  • $\begingroup$ Thanks Did, this is very nice and makes it quite manageable! Just a clarification on it. How do you go from $p(Y>X_1-X_2+\pi, X_2>\pi)$ to having $X_2-X_1-\pi$ inside $\Phi$ which I'm assuming is the cumulative distribution function? Secondly, if I also had the condition $X_1>X_2-\pi$, then would I just integrate $X_1$ first from $X_2-\pi$ to $\inf$ and $X_2$ from $\pi$ to $\inf$? Thanks again for your help! $\endgroup$ – Hakeem May 26 '14 at 7:22
  • $\begingroup$ $[Y\gt X_1-X_2+\pi]=[Z\lt(X_2-X_1-\pi)/\sigma_Y]$ where $Z=-Y/\sigma_Y$ is standard normal. $\endgroup$ – Did May 26 '14 at 7:26
  • $\begingroup$ Thanks for clearing that up! You're so quick, so you may have missed the second question I was editing in, so I am reposting. If I also had the condition $X_1>X_2-\pi$, then would I just integrate $X_1$ first from $X_2-\pi$ to $\inf$ and $X_2$ from $\pi$ to $\inf$? $\endgroup$ – Hakeem May 26 '14 at 7:41
  • $\begingroup$ Yes. $ $ $ $ $ $ $\endgroup$ – Did May 26 '14 at 7:45
  • $\begingroup$ Thanks for all your help, appreciate it! $\endgroup$ – Hakeem May 26 '14 at 8:28

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.