I am stuck with a question doing one of my stats tutorial and question is as follows:

Suppose X and Y are two independent exponential random variables with parameter $\theta$, i.e. their joint probability density function is

$f(x,y; \theta) = \frac{1}{\theta^2}e^\frac{-(x+y)}\theta, x\geq0, y\geq0$

where parameter $\theta >0$.

Find the probability density function of $Z = \frac{X}{X+Y}$

Can anyone kindly guide me with this question please? I'm not exactly sure how to begin. Thank you!

  • 1
    $\begingroup$ The section 'Dependent variables and change of variables' in the Wikepedia article about probability density functions will be helpful. $\endgroup$ – Nussig Sep 15 '18 at 10:14
  • 1
    $\begingroup$ You might want to add the self-study tag and read it's wiki. $\endgroup$ – StubbornAtom Sep 15 '18 at 10:17

The joint density of $(X,Y)$ is of the form

\begin{align} f_{X,Y}(x,y)&=\frac{1}{\theta^2}\exp\left({-\frac{x+y}{\theta}}\right)\mathbf1_{x,y>0}\quad,\,\theta>0 \\&=\frac{1}{\theta}e^{-x/\theta}\mathbf1_{x>0}\frac{1}{\theta}e^{-y/\theta}\mathbf1_{y>0} \\&=f_X(x)f_Y(y)\quad,\text{ say } \end{align}

So $X$ and $Y$ are independent and identically distributed Exponential variables with mean $\theta$.

You seek the distribution of $\frac{X}{X+Y}=Z$ (say).

Among several ways to find the distribution of $Z$, we could find the distribution function (DF) $P(Z\leqslant z)$ of $Z$, or we could use a change of variables (as mentioned in a comment) along the lines of $(X,Y)\to (Z,W)$ such that $Z=\frac{X}{X+Y}$ and $W=X+Y$.

For the DF, we can proceed using the total probability theorem :

For each $z\in(0,1)$,

\begin{align} F_Z(z)&=P\left(\frac{X}{X+Y}\leqslant z\right) \\&=\int P(X\leqslant zX+y\mid Y=y)f_Y(y)\,dy \\&=\int_0^{\infty}P\left (X\leqslant \frac{y}{1-z}\right)f_Y(y)\,dy \\&=\cdots \end{align}

Differentiating $F_Z$ wrt $z$ would yield the density of $Z$.

If you use the second method with that particular transformation, you would find from the joint density of $(Z,W)$ that $Z$ and $W$ are also independently distributed, and finally identify the distribution of $Z$ from the joint density alone.

  • $\begingroup$ Thank you very much for your answer! This was the solution I was seeking! @StubbornAtom $\endgroup$ – InvadersMustDie Sep 17 '18 at 16:24
  • $\begingroup$ @InvadersMustDie Let me know your final answer. $\endgroup$ – StubbornAtom Sep 17 '18 at 19:07
  • 1
    $\begingroup$ it is a uniform distribution U, since the pdf = 1 $\endgroup$ – InvadersMustDie Sep 18 '18 at 18:28

A common method is to find CDF, then differentiate: $$F_Z(z)=P(Z \leq z) = P(\frac{X}{X+Y}\leq z) = P(\frac{1-z}{z}X\leq Y)$$ which is 1 outside $0 \leq z \leq 1$.Let $\alpha=\frac{1-z}{z}$, and we seek for $P(\alpha X\leq Y)$, where $\alpha \geq 0$. In the 2D plane, you'll draw the line $y=\alpha x$, and integrate the joint PDF in the area in-between the line and the x-axis; then substitute $z$, and differentiate with respect to it.


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.