I am interested in finding an analytic expression for the length of a 3-vector whose components are distributed according to a Laplace distribution with zero mean and the same scale parameter.

I have been unable to find anything useful so far in the literature, but it seems from my numeric tests that the answer should be close to $r^2 e^{-r/\lambda} dr$, where $r = \sqrt{x_1^2 + x_2^2 + x_3^2}$ and the $x_i$ are Laplace variables with zero mean and exponential scale parameter $\lambda$. My heuristic result makes some sense to me from a transformation to spherical coordinates.

It would already help if I knew something more about the distribution of $r^2$ (for normally-distributed $x_i$, I know that I would get a $\chi^2$ distribution).

I would appreciate any hints or literature pointers you might have.


  • $\begingroup$ Length of a vector is generally considered to be a nonnegative quantity and so why are you modeling it as a Laplacian radio variable rather than an exponential random variable? $\endgroup$ – Dilip Sarwate Feb 4 '19 at 22:38
  • $\begingroup$ Exponential distributions are perfectly fine, I would appreciate any input in that regard as well. It is just that my physical problem starts out with a Laplace distribution (these are real-space vectors pointing in all directions). $\endgroup$ – sblatt Feb 4 '19 at 22:48
  • $\begingroup$ Are the components independent? $\endgroup$ – kjetil b halvorsen Feb 4 '19 at 23:04
  • $\begingroup$ Yes. They are independent, but drawn from the same distribution. $\endgroup$ – sblatt Feb 4 '19 at 23:21

Consider $X_1, X_2, X_3\sim\text{Laplace}(0,b)$, iid for a parameter $b$.

Then $|X_i|\sim\text{Exp}\big(\frac{1}{b}\big)$.

Then $X_i^2=|X_i|^2\sim\text{Weibull}\big(b^2,\frac{1}{2}\big)$.

So your question is answered by the sum of three iid Weibull variables with common scale $b^2$ and the specific shape $\frac{1}{2}$.

Unfortunately, there does not seem to be anything on the sum of iid Weibulls in general, and convolving even only two Weibulls with shape $\frac{1}{2}$ exceeds WolframAlpha's standard computing time.

| cite | improve this answer | |

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.