Let X, Y have joint pdf: $f(x,y) = \frac{1}{2}(x+y)e^{(-x+y)}$ for $0<x<\infty$ and $0<y<\infty$.

Find the probability density function $f_Z(z)$ for the sum $Z = X+Y$ for all $-\infty<z<\infty$.

I am trying to solve it by finding the marginal distributions $f_x(x)$ and $f_y(y)$, then integrating $\int_{-\infty}^{\infty}f_x(z-y)f_ydy$ to get $f_z(z)$. Is this the right approach?

  • $\begingroup$ Your pdf is plainly incorrect in the term $exp(-x+y)$, as this will explode out for large $y$. Presumably you mean $exp[-(x+y)]$ ... $\endgroup$ – wolfies Dec 5 '13 at 12:40

Going from the marginals will in general not work, for the marginals forget about dependence.

The joint density lives in the first quadrant, so $Z\gt 0$. Thus we are only interested in the density of $Z$ for $z\gt 0$.

I would travel via the cdf of $Z$. We want the probability that $X+Y\le z$. For that, we want to integrate $(x+y)e^{-(x+y)}$ over the part of the first quadrant below the line $x+y=z$. Because of the shape of the integrand, it is best to use a transformation, $x+y=u$, $y=v$. (It is somewhat more pleasant to find $\Pr(Z\gt z)$.)

Remark: The following informal idea can be turned into a formal argument. The density function $f_Z(z)$, times $dz$, is roughly the probability that $Z$ lies between $z$ and $z+dz$. Looking at the density function, we see this is approximately $ze^{-z} dz$.


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.