PDF of the sum distribution of a Laplace and a Gamma distributions

249 Views Asked by At

Let $X,Y$ be independent randomly distributed continuous variables with the following PDF

$f_X(x) = \frac{\lambda}{2}e^{-\lambda\mid x \mid}$ for $x \in \mathbb{R}$

$f_Y(y) = \lambda^2ye^{-\lambda y}$

What's the PDF of the sum distribution $Z = X + Y$?

My working:

I've found out that $X \sim Laplace(0,\frac{1}{\lambda})$ and $Y \sim Gamma(2,\frac{1}{\lambda})$.

I'm willing to apply the following formula (which I think is the correct way to go):

$f_Z(z) = \int_\mathbb{R} f_X(x)f_Y(z-x) dx$.

So I calculate $f_X(x)f_Y(z-x)$ (using MATLAB) and I get

$f_X(x)f_Y(z-x) = -\frac{1}{2}(\lambda^3e^{-\lambda\mid x \mid}e^{\lambda(x-z)}(x - z)$.

Now, when I integrate with respect to x, I dont get any sensible results, and I assume that's because I'm not getting the integration limits correctly.

How are the appropiate integration limits found?

1

There are 1 best solutions below

4
On

long calculations to me (I am not familiar with long calculations) but easy exercise:

Set $Z=X+Y$ and $U=Y$. After verifying that Jacobian is 1, the joint density of $(U,Z)$ is

$$f_{UZ}(u,z)=\frac{\lambda^3}{2}ue^{-\lambda u}e^{-\lambda|z-u|}$$

Now to get $f_Z(z)$ you have to integrate in $du$. To do that observe that:

IF $Z<0$ the expression inside the module is always $<0$ thus

$$f_{UZ}(u,z)=\frac{\lambda^3}{2}e^{\lambda z}\cdot ue^{-2\lambda u}$$

and

$$f_Z(z)=\frac{\lambda^3}{2}e^{\lambda z}\int_0^{\infty} ue^{-2\lambda u}du=\frac{\lambda}{8}e^{\lambda z}$$

IF $Z>0$ you have to integrate accordingly splitting the module...

After some easy but tedious calculations the result is the following:

$$ f_Z(z) = \begin{cases} \frac{\lambda}{8}e^{\lambda z}, & \text{if $z<0$ } \\ \frac{\lambda^3}{4}e^{-\lambda z}\left[z^2+\frac{2\lambda z+1}{2\lambda^2}\right], & \text{if $z\geq 0$ } \end{cases}$$


Just as an example, setting $\lambda=1$ this is the resulting density

enter image description here