2
$\begingroup$

I'm trying to show that for uniformly distributed variables $X_1$ and $X_2$ that the vector $$ (\sqrt{-2\log X_1} \cos(2\pi X_2), \sqrt{-2 \log X_1} \sin(2\pi X_2)) $$ is two-dimensional normal. Letting $Y_1$ be the first component and $Y_2$ be the second component I see that $X_1 = e^{-(Y_1^2 +Y_2^2)/2}$ and $X_2 = \frac{1}{2\pi} \arctan(Y_2/Y_1)$. Is there a way to proceed from here to show that $(Y_1, Y_2)$ is normal?

$\endgroup$
1
  • $\begingroup$ Hint: do a transformation from cartesian to polar coordinates. $\endgroup$
    – orion
    Nov 14 '14 at 18:49
2
$\begingroup$

Remember the substitution rule (if we can characterize the CDF, then we are done). In particular, consider the absolute value of the determinant of the Jacobian matrix \begin{align*} \left|\det\left[\begin{array}{cc} \frac{\partial X_{1}}{\partial Y_{1}} & \frac{\partial X_{1}}{\partial Y_{2}}\\ \frac{\partial X_{2}}{\partial Y_{1}} & \frac{\partial X_{2}}{\partial Y_{2}} \end{array}\right]\right| & =\left|\det\left[\begin{array}{cc} -Y_{1}\exp\left(Y_{1}^{2}/2-Y_{2}^{2}/2\right) & -Y_{2}\exp\left(Y_{1}^{2}/2-Y_{2}^{2}/2\right)\\ -Y_{2}/\left[2\pi\left(Y_{1}^{2}+Y_{2}^{2}\right)\right] & Y_{1}/\left[2\pi\left(Y_{1}^{2}+Y_{2}^{2}\right)\right] \end{array}\right]\right|\\ & =\left|-Y_{1}^{2}\frac{\exp\left(Y_{1}^{2}/2-Y_{2}^{2}/2\right)}{2\pi\left(Y_{1}^{2}+Y_{2}^{2}\right)}-Y_{2}^{2}\frac{\exp\left(Y_{1}^{2}/2-Y_{2}^{2}/2\right)}{2\pi\left(Y_{1}^{2}+Y_{2}^{2}\right)}\right|\\ & =\left|-\frac{\exp\left(Y_{1}^{2}/2-Y_{2}^{2}/2\right)}{2\pi}\right|\\ & =\frac{\exp\left(Y_{1}^{2}/2\right)}{\sqrt{2\pi}}\frac{\exp\left(Y_{2}^{2}/2\right)}{\sqrt{2\pi}}. \end{align*}

$\endgroup$

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.