Let $(X,Y)$ have the mixed discrete-continuous pdf given by:

$$f(x,y)= \begin{cases} \frac{y^{a+x-1}e^{-2y}}{\Gamma(a) x!}\ y>0;x=0,1,2,\ldots \\ 0 \ \ \ \ \ \ \ \ \ \ \ \ \ \ \text{elsewhere} \end{cases}$$

Could you please help me show that this nasty pdf integrates/sums to 1 over the support of $(X,Y)$?

I initially tried to integrate out $Y$, by completing the gamma distribution with parameters, shape**$=a+x$ and **scale=$1/2$ and then sum over $X$ but that lead to a mess.

Is there another way to go here? Thank you.

  • $\begingroup$ I see a $z$ in your nasty pdf! $\endgroup$ – Stat May 30 '14 at 16:24
  • $\begingroup$ @Stat It's an $x$, sorry about that. $\endgroup$ – JohnK May 30 '14 at 16:25

Really fascinating question. What you want to show is that

$$\int\limits_{0}^{\infty}\sum\limits_{x=0}^{\infty}\dfrac{y^{a+x-1}e^{-2y}}{\Gamma\left(a\right)x!}\text{ d}y = 1$$

Here's how I approached this problem. Let's isolate the $x$ and $y$ parts as much as possible:

$$\int\limits_{0}^{\infty}\dfrac{y^{a-1}e^{-2y}}{\Gamma\left(a\right)}\left(\sum\limits_{x=0}^{\infty}\dfrac{y^{x}}{x!}\right)\text{ d}y$$

I thought to myself, well, $y > 0$ and the $\displaystyle \sum\limits_{x=0}^{\infty}\dfrac{y^{x}}{x!}$ part looks something like the CDF of a Poission random variable. But the probability mass function of a Poission random variable with mean $\lambda = y$ is $$ p\left(y\right) = \dfrac{e^{-y}y^{x}}{x!}\text{,} \quad{} y > 0\text{.}$$

So this suggested to me that I should bring one of the $e^{-y}$ terms into the summation.

$$\int\limits_{0}^{\infty}\dfrac{y^{a-1}e^{-y}}{\Gamma\left(a\right)}\left(\sum\limits_{x=0}^{\infty}\dfrac{e^{-y}y^{x}}{x!}\right)\text{ d}y$$

Since I've already established that the summation is summing over all probabilities for a Poisson-distributed random variable, it follows that $\displaystyle \sum\limits_{x=0}^{\infty}\dfrac{e^{-y}y^{x}}{x!} = 1$.

So all that remains is to show that

$$\int\limits_{0}^{\infty}\dfrac{y^{a-1}e^{-y}}{\Gamma\left(a\right)}\text{ d}y = 1\text{.}$$

However, this follows immediately since $\displaystyle f(y) = \dfrac{y^{a-1}e^{-y}}{\Gamma\left(a\right)}$ is the PDF of a Gamma distribution.

  • 1
    $\begingroup$ Very good, thank you. Also $e^x=\sum_{i=0}^{\infty} \frac{x^n}{n!}$ ;) $\endgroup$ – JohnK May 30 '14 at 19:00
  • $\begingroup$ I was thinking about that, but for some reason, I thought it was $\displaystyle e^{x} = \sum\limits_{i = 1}^{\infty}\dfrac{n^{x}}{n!}$ when I wrote it at the time! Of course, that's not correct. :P $\endgroup$ – Clarinetist May 30 '14 at 19:02

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.