I'm wondering if there's a distribution $\Gamma$ such that if we draw $x_i \dots d_n$ iid from $\Gamma$, then $\sum_{i=1}^n |x_i|$ has a nice distribution. I thought maybe the normal distribution, since then the absolute value would be half-normal, but I'm not sure what the distribution of the sum of half-normal variables are.

  • 1
    $\begingroup$ If you're willing to consider a distribution that's supported on the positive reals to begin with, there's a nice expression for the distribution of a sum of independent exponential random variables (en.wikipedia.org/wiki/Erlang_distribution) $\endgroup$ – Julian Rosen Sep 20 '12 at 4:18

If i.i.d. random variables $X_1, X_2, \ldots, X_n$ have Laplacian density $$f(x) = \frac{1}{2}\exp\left(-|x|\right), -\infty < x < \infty,$$ then $|X_i|$ are i.i.d. exponential random variables with mean $1$, and $\sum_{i=1}^n |X_i|$ is a Gamma random variable with mean $n$ and order parameter $n$. This is effectively the same result as the one in the comment by @PinkElephants but without the requirement of support on the positive reals only (which makes the distinction between $X_i$ and $|X_i|$ irrelevant.)

  • $\begingroup$ I think the cases given thus far all satisfy the conditions for the central limit theorem. So for large n there is a normal distribution approximation that should be pretty good and can be normalized so that a standard normal approximation can be used. $\endgroup$ – Michael R. Chernick Sep 20 '12 at 19:48

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.