Re: clustering vs. random
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Tue, 11 Jul 2006 22:13:46 -0400
On 11 Jul 2006 15:45:18 -0700, hhdave@xxxxxxx wrote:
Let's say there's a street with 150 houses on it. If I want to
determine whether the 25 red houses were clustered together, as opposed
to distributed randomly among the entire street, what would be the best
test for this?
Expanding this, how do we determine whether houses tend to be clustered
according to color?
Assume that the houses are not equally spaced along the street; the
next-door distances are log-normal distributed.
Thanks for any insight!
Look up < disease clustering epidemiology >
I don't know what the assumption of 'log-normal' distances
does, unless it says that you need a Monte-Carlo type solution.
One metric used for disease clusters, which is often based
on houses, is the reciprocal of the distance D. The bigger sum
for D^-1 has more smaller distances.
So you could take 10,000 samples of 25 houses from the
150 houses, and compute the total of D^-1 for each pair,
for the 25*24/2 pairs of houses; and compare that distribution
to the sum for 25 red houses.
--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.
- References:
- clustering vs. random
- From: hhdave
- clustering vs. random
- Prev by Date: clustering vs. random
- Next by Date: plot satellite data on map
- Previous by thread: clustering vs. random
- Next by thread: plot satellite data on map
- Index(es):
Relevant Pages
|
