Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

In a square with dimension $M$ I have a fixed point $A$ with $(x_a,y_a)$ coordinates and some other random nodes $B_1,B_2,\dots,B_n$ with random uniformly distributed coordinates. I want to find out the mean distance between a fixed point and a uniformly distributed random variable.

share|improve this question
add comment

2 Answers

It seems one does not need to introduce more than one uniformly distributed random point $B$ and that you look for $m(A)=E(\text{dist}(A,B))$. Assuming without loss of generality that the square has side $1$, $$ m(x_a,y_a)=\int_0^1\int_0^1\sqrt{(x-x_a)^2+(y-y_a)^2}\mathrm{d}x\mathrm{d}y. $$ One can write down several equivalent formulas for $m(x_a,y_a)$, none of them very enlightening, and I doubt that a general expression exists, involving only usual functions.

Nevertheless, obvious symmetries of the problem yield that $m$ is invariant under the action of the transformations $(x_a,y_a)\mapsto(y_a,x_a)$, $(x_a,y_a)\mapsto(1-x_a,y_a)$ and $(x_a,y_a)\mapsto(x_a,1-y_a)$. And the function $m$ is maximal at the four corners and minimal at the center of the square.

One can compute the value of $m$ at these points. For example, using polar coordinates $(r,\theta)$, $$ m(0,0)=2\int_0^{\pi/4}\int_0^{1/\cos\theta}r^2\mathrm{d}r\mathrm{d}\theta=\frac23\int_0^{\pi/4}\frac{\mathrm{d}\theta}{\cos^3\theta}=\frac23\int_0^{\sqrt2/2}\frac{\mathrm{d}s}{(1-s^2)^2}, $$ which should yield something like $ m(0,0)=\frac13(\sqrt2+\log(1+\sqrt2))=0.765195716. $

And if $A$ is the center of the square, $m(A)$ is also the mean distance between $A$ and a random point $B'$ uniformly distributed in the square $\frac12\le x,y\le1$. The point $A$ and this smaller square are homothetic to the point $(0,0)$ and the original square, with ratio $\frac12$, hence $ m(\frac12,\frac12)=\frac12m(0,0). $

share|improve this answer
add comment

The distance of $(x_a, y_a)$ to $B_i=(x_i,y_i)$ is $d_i=\sqrt{ (x_a-x_i)^2+(y_a-y_i)^2 }$

You want to find $(1/n) \sum_i \mathrm E [d_i]$

Are you familiar with transformations of random variables?

share|improve this answer
    
Yes, I want to find out if this integral is calculated before? –  Mehdi Jul 11 '11 at 6:10
    
Also mention that I want to find out the mean in general so I use the keyword "integral" instead "summation" –  Mehdi Jul 11 '11 at 6:11
    
So the number of random variables is not countable? In any case you need to calculate the expectation first. Can you do that? –  Emre Jul 11 '11 at 16:05
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.