Let $X$ and $Y$ be independent exponential variables with rates $\alpha$ and $\beta$, respectively. Find the CDF of $X/Y$.

I tried out the problem, and wanted to check to see if my answer of: $\frac{\alpha}{ \beta/t + \alpha}$ is correct, where $t$ is the time, which we need in our final answer since we need a cdf.

Can someone verify if this is correct?


Recall one of the most important characterizations of the exponential distribution:

The random variable $Y$ is exponentially distributed with rate $\beta$ if and only if $P(Y\geqslant y)=\mathrm{e}^{-\beta y}$ for every $y\geqslant0$.

Let $Z=X/Y$ and $t\gt0$. Conditioning on $X$ and applying our characterization to $y=X/t$, one gets $$ P(Z\leqslant t)=P(Y\geqslant X/t)=E(\mathrm{e}^{-\beta X/t}). $$ Now, the density of the distribution of $X$ is $\alpha\mathrm{e}^{-\alpha x}$ on $x\geqslant0$, hence for every $\gamma\geqslant0$, $$ E(\mathrm{e}^{-\gamma X})=\int_0^{+\infty}\alpha\mathrm{e}^{-(\alpha+\gamma) x}\mathrm{d}x=\frac{\alpha}{\alpha+\gamma}\left[-\mathrm{e}^{-(\alpha+\gamma) x}\right]_{0}^{+\infty}=\frac{\alpha}{\alpha+\gamma}. $$ Substituting $\gamma=\beta/t$ yields the formula.

  • $\begingroup$ What is the probability distribution function? $\endgroup$ – user Mar 11 '14 at 4:37
  • 4
    $\begingroup$ @user How to deduce the PDF from the CDF? Tell me... $\endgroup$ – Did Mar 11 '14 at 6:34
  • $\begingroup$ @Did how did you arrive to $P(Y\geqslant X/t)=E(\mathrm{e}^{-\beta X/t})$. How is the probability and expectation related? $\endgroup$ – user35443 Nov 1 '18 at 13:57
  • $\begingroup$ @user35443 For every $x\geqslant0$, $P(Y\geqslant x/t)=e^{-\beta x/t}$. Now, $X$ is independent of $Y$ hence $P(Y\geqslant X/t\mid X)=e^{-\beta X/t}$ almost surely. Finally, take expectations on both sides. $\endgroup$ – Did Nov 3 '18 at 14:56
  • $\begingroup$ Makes sense, thanks! $\endgroup$ – user35443 Nov 4 '18 at 18:18

Here's a slightly different point of view. \begin{align} \Pr\left( \frac X Y \ge t \right) & = \iint\limits_{\{\,(x,y)\,:\, x\,\ge\,ty\,\ge\,0 \,\}} e^{-\alpha x} e^{-\beta y} (\alpha\beta\,d(x,y)) \\[10pt] & = \int_0^\infty \left( \int_{ty}^\infty e^{-\alpha x} (\alpha\,dx) \right) e^{-\beta y} (\beta\,dy) \\[10pt] & = \int_0^\infty (e^{-\alpha ty}) e^{-\beta y} (\beta\,dy) \\[10pt] & = \beta \int_0^\infty e^{-(\alpha t+\beta)y} \, dy = \frac \beta {\alpha t + \beta}. \end{align} This is $1$ minus the c.d.f. Find the c.d.f. and differentiate to get the p.d.f. on the interval $t\ge0.$


This is correct. I did the calculation and got the same answer.

  • $\begingroup$ Did you use u-substitutions, a lot of them? $\endgroup$ – mary Apr 19 '11 at 5:02
  • $\begingroup$ No, I used Mathematica :) If you do it by hand and you are not familiar with $\int \alpha e^{-\alpha x} dx$ you might need a lot of substitutions. $\endgroup$ – GWu Apr 19 '11 at 5:03

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.