> Maybe it would make sense to compute, instead of a mean, some kind of
> "weighted" location, (like they do when mixing "more than two"
> navigational aid signals e.g.)  (well, it is done for VOR and DME - I
> am not sure about GPS...) where the weight given to each signal, is
> inversely proportional to some power, like the square, of the
> distance away from where the "weighted location" would be, "without"
> that input datum ("data item"). ((I guess you'd have to use a mean or
> something to get "started"...))

I don't have to use the mean as a starting point.  I'm using a kmeans
analysis of the users that I've been playing with as I've added users.
Kmeans starts with an idea of a certain, prespecified number of clusters
and finds the fit of the data points that clusters them together into
clusters of least distance from the core.

It isn't easy to see, but two of the markers (one red and one yellow) 
have a number in the icon.  Those are the center of the clusters.  The 
rest are the user data points and the color corresponds to which cluster 
the user belongs to.

http://plug.phoenix.az.us/membermap/test.html
---------------------------------------------------
PLUG-discuss mailing list - PLUG-discuss@lists.plug.phoenix.az.us
To subscribe, unsubscribe, or to change your mail settings:
http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss