0
$\begingroup$

Let X,Y be independent random variables following $N(0,1)$. What is the probability $P(X \geq 0, X+Y \geq 0)$?

I know this will be $\displaystyle\int_0^{\infty}\int_{-x}^{\infty}\frac{1}{2\pi}e^{-\frac{x^2+y^2}{2}}dydx$ the problem is I cannot evaluate the latter. Can someone help please?

$\endgroup$
  • $\begingroup$ there is NO closed expression integration over 'y' gives error function $\endgroup$ – Jose Garcia Feb 7 '17 at 10:27
  • $\begingroup$ @JoseGarcia Doesn't mean $\int_0^\infty \Phi(x)e^{-x^2/2}$ doesn't have a closed form expression. $\endgroup$ – spaceisdarkgreen Feb 7 '17 at 10:32
  • $\begingroup$ I know that, but the double integral should be computable with some trick. This comes from having a random Gaussian walk and asking the probability that both x_1 and x_2 are positive. $\endgroup$ – Ted Feb 7 '17 at 10:32
  • $\begingroup$ mathforum.org/library/drmath/view/52137.html - see the third from last integral $\endgroup$ – Cato Feb 7 '17 at 10:32
  • $\begingroup$ Oh, great, thanks Cato! Polar coordinates do the trick! $\endgroup$ – Ted Feb 7 '17 at 10:36
2
$\begingroup$

We need to compute $$\int_0^\infty \Phi(x) \frac{1}{\sqrt{2\pi}}e^{-x^2/2}dx$$ where $\Phi(x) = \int_{-\infty}^x e^{-t^2/2}dt/\sqrt{2\pi}=\int_{-x}^\infty e^{-t^2/2}dt/\sqrt{2\pi}$ is the normal cumulative distribution function.

Since $\Phi'(x) = e^{-x^2/2}/\sqrt{2\pi},$ this can be rewritten $$ \int_0^\infty\Phi(x)\Phi'(x)dx= \int_{1/2}^1\Phi d\Phi = 3/8 $$ where we changed variables to $\Phi = \Phi(x).$

More generally if you want the probability that $a < X < b$ and $X+Y>0,$ you get $$\int_a^b\Phi(x)\Phi'(x)dx = \frac{1}{2}(\Phi(b)^2-\Phi(a)^2).$$

$\endgroup$
0
$\begingroup$

Using polar coordinates and thinking about the shape of the domain we get : $$ \frac{1}{2\pi} \int_0^\infty \int_{-x}^\infty \exp(\frac{-x^2-y^2}{2}) \ dx dy = \frac{1}{2\pi} \int_{-\pi/4}^{\pi/2} d\phi\ \int_0^\infty dr\ \exp(-r^2/2)\cdot r = $$ $$ \frac{\pi/2 + \pi/4}{2\pi} \int_0^\infty \exp(-r^2/2)\cdot r\ dr = \frac{\pi/2 + \pi/4}{2\pi} \cdot 1 = \frac{3}{8} $$

$\endgroup$
  • $\begingroup$ should be $-\pi/4$ to $\pi/2.$ You just caught the sign error too, $\endgroup$ – spaceisdarkgreen Feb 7 '17 at 10:51
  • $\begingroup$ Almost... now double check $\int_0^\infty e^{-r^2/2}rdr$ $\endgroup$ – spaceisdarkgreen Feb 7 '17 at 10:54
  • $\begingroup$ My God this was tedious... I should go back sleeping. Thank you and sorry for the mistakes. $\endgroup$ – Zubzub Feb 7 '17 at 10:56
  • 1
    $\begingroup$ haha no problem, just need to make sure two correct methods give the same answer. $\endgroup$ – spaceisdarkgreen Feb 7 '17 at 10:56

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.