1
$\begingroup$

I am not getting where i am doing mistake in deriving the PDF of ratio of two normalrandom variables. Here is what i had did:

Say we have two random variables $X$ and $H$ with PDF as $f_{X}(x) = \frac{1}{\pi \sigma_x}K_0\left(\frac{|x|}{\sigma_x}\right)$ and $f_H(h)=\frac{1}{\sqrt{2\pi}\sigma_h}e^{-\frac{h^2}{2\sigma_h^2}}$. Then using formula for PDF of ratio of two random variables from M.D.Springer book(P.No.92), the PDF of $Z=\frac{\alpha X}{H}$ is obtained as

$f_Z(z)=\int_{-\infty}^{+\infty}h \frac{1}{\pi \sigma_x}K_0\left(\frac{|hz|}{\alpha\sigma_x}\right)\frac{1}{\sqrt{2\pi}\sigma_h}e^{-\frac{h^2}{2\sigma_h^2}}dh$-------(1)

However in the research paper (which i am refereeing for the derivation), the PDFof $Z$ is given as $f_Z(z)=\int_{-\infty}^{+\infty}|h| \frac{1}{\alpha\pi \sigma_x}K_0\left(\frac{|hz|}{\alpha\sigma_x}\right)\frac{1}{\sqrt{2\pi}\sigma_h}e^{-\frac{h^2}{2\sigma_h^2}}dh$------(2)

I am not getting how the equation (2) is coming. Any help in this regard is highly appreciated.

$\endgroup$
0
1
$\begingroup$

Let $Z=\alpha X/H$ with $\alpha>0$. Then $$ \begin{aligned} \mathsf P(Z\leq z)% &=\mathsf P(\alpha X/H\leq z)\\ &=\mathsf P(\alpha X/H\leq z,H>0)+\mathsf P(\alpha X/H\leq z,H<0)\\ &=\mathsf P(X\leq zH/\alpha,H>0)+\mathsf P(X\geq zH/\alpha,H<0). \end{aligned} $$ Assuming $X$ is independent of $H$ we then have $$ \begin{aligned} \mathsf P(Z\leq z)% &=\int_0^\infty\int_{-\infty}^{zh/\alpha}f_X(x)f_H(h)\,\mathrm dx\mathrm dh+\int_{-\infty}^0\int_{zh/\alpha}^\infty f_X(x)f_H(h)\,\mathrm dx\mathrm dh. \end{aligned} $$ Differentiating we have by the Leibniz integral rule: $$ \begin{aligned} f_Z(z)=\partial_z\mathsf P(Z\leq z)% &=\int_0^\infty \frac{h}{\alpha}f_X(zh/\alpha)f_H(h)\,\mathrm dh-\int_{-\infty}^0 \frac{h}{\alpha} f_X(zh/\alpha)f_H(h)\,\mathrm dh\\ &=\int_{-\infty}^\infty \frac{|h|}{\alpha}f_X(zh/\alpha)f_H(h)\,\mathrm dh. \end{aligned} $$ Substituting in the appropriate quantities for $f_X$ and $f_H$ then gives $(2)$.

$\endgroup$

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.