3
$\begingroup$

The result for the average distance between 2 points with normally distributed coordinates has already been demonstrated on this site and I found a white paper for the generalized result when these 2 points have $N$ dimensions.

But I am at a loss to compute how the average distance changes as the number of points increases to some $M > 2$. Actually, the ultimate goal is to determine the ratio of the average distance between $M_1$ number of points and the average distance between $M_2$ number of different points (all taken from the same normal distribution).

Ideally the result would be generalized to $N$ dimensions, but if I'm asking for too much, I'd be more than happy to learn of the 2D result.

Thanks for any help...

EDIT: To clarify my request, if $M = 3$, the average distance between the 3 points with (random) normally distributed coordinates would be: $$[\textrm{dist}(p_1, p_2) + \textrm{dist}(p_1, p_3) + \textrm{dist}(p_2, p_3)] / 3$$ However, if an approximation to the ratio mentioned in paragraph 2 can be found using a statistical approach (perhaps the ratio of two confidence intervals, for $M_1$ and $M_2$ respectively, that the centroid is at the origin), that would work for me. Thanks, Sasha, for this idea.

$\endgroup$
2
  • $\begingroup$ How do you define the average distance? Distance between centroids of two data sets? Mean of pair-wise distances? Something else? $\endgroup$
    – Sasha
    Nov 2, 2012 at 20:21
  • $\begingroup$ I was aiming for the mean of pair-wise distances. Forgive my ignorance, but wouldn't the centroids of the two data sets be the same on average, regardless of number of points? $\endgroup$
    – Kevin
    Nov 2, 2012 at 20:37

1 Answer 1

Reset to default
2
$\begingroup$

By linearity of expectation, the mean of the pairwise distances is just the average distance between two points.

$\endgroup$
1
  • $\begingroup$ Embarrassingly for me, you are correct. Thanks. $\endgroup$
    – Kevin
    Nov 2, 2012 at 22:34

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.