Suppose X,Y are iid normal distribution with $\mu = 0$, what is the distribution of ratio of absolute value of X to Y?

36 Views Asked by At

Let X and Y be iid normal distribution with mean $\mu = 0$, let $Z=\frac{|X|}{|Y|}$, what is the distribution of Z?

I'm thinking of using the distribution method but I'm stuck by the integration. I'm also wondering if change of variable technique can be used.

1

There are 1 best solutions below

1
On BEST ANSWER

I'm not sure if it has a name; however, the ratio $X/Y$ itself has a well-known, named distribution, so you can just take the absolute value of that random variable.


First off, $X/Y = (X/\sigma)/(Y/\sigma)$, so we may assume without loss of generality that $X,Y$ are standard normal random variables (with mean $0$ and standard deviation $1$)

Then we may write $$\mathbb{P}(X/Y \leq 0) = \mathbb{P}(X\leq 0, Y>0) + \mathbb{P}(X\geq 0, Y<0) = \frac12,$$ so for $r > 0$, $$\mathbb{P}(X/Y \leq r) = \frac12 + \mathbb{P}(0 < X/Y \leq r) = \frac12 + 2\mathbb{P}(X\leq rY, X>0, Y>0).$$ This last expression can be found by a polar substitution: $$\begin{align*}\mathbb{P}(X \leq rY, X>0, Y>0) &= \frac{1}{2\pi}\int_0^\infty\int_0^{ry}e^{-\tfrac12(x^2+y^2)}\,dx\,dy \\ &= \frac1{2\pi}\int_0^\infty\int_0^{\arctan(r)}\rho e^{-\rho^2/2}\,d\theta\,d\rho \\ &= \frac1{2\pi}\arctan(r)\end{align*}$$ giving $$\mathbb{P}(X/Y \leq r) = \frac12 + \frac1\pi\arctan(r).$$

The expression for $r < 0$ can be found by symmetry as $\mathbb{P}(X/Y \leq r) = \mathbb{P}(X/Y \geq -r)$, which gives the same expression as above, just extended to all real $r$.