# Average distance between random points inside a cube

Let $U_1, U_2, \ldots, U_n$ be a set of random variable uniformly distributed over some box in $\mathbb{R}^3$ and let $$R = \frac{1}{2 n(n-1)}\sum\limits_{i,j< i} |U_i - U_j|$$ be the random variable corresponding to the average distance between the random points. What is the distribution of $R$? If no answer to that is available, then what is $E(R)$?

I have found somewhat similar questions, but couldn't related them to this one.

Thank you very much in advance.

Gabriel

-
E(R) is rather clumsy to compute: math.utep.edu/Faculty/moschopoulos/Publications/… –  leonbloy Jan 14 '13 at 15:30
add comment

## 1 Answer

I take it that the $U_i$ are independent. As $\binom{n}{2}=n(n-1)/2$ I assume that you mean to investigate $$R=\frac{2}{n(n-1)}\sum_{j<i}|U_i-U_j|.$$ The distribution of $R$ is not any of the well-known distributions. By linearity, $E[R]=E[|U_i-U_j|]$ which you can find by integration over $\text{box}\times\text{box}$, i.e. $$E[R]=\frac{1}{\operatorname{vol}(\text{box})^2}\int_{\text{box}\times\text{box}}|x-y|\,d^3x\,d^3y.$$

-
I am still confused about where the concentration of points enter. I mean, if $n$ is large, than the average distance between them should be smaller, but I don't see how that enters the calculation. Anyway, thank you for the help. –  Gabriel Landi Jan 14 '13 at 15:30
"I mean, if n is large, than the average distance between them should be smaller" No. YOu are perhaps thinking of the minimum distance -the average distance to the nearest neighbour- which is another thing. –  leonbloy Jan 14 '13 at 15:32
Oh! Yeah, I think that is it. The nearest-neighbour distance. I got everything confused. –  Gabriel Landi Jan 14 '13 at 15:33
add comment