Brian - thanks for the suggestions but I'm not sure how the Mahalanobis distance would apply in this case. The regular Euclidean distance formula is really the same as the distance which is part of the polar co-ordinate but lacks the useful angle information.
One thing I didn't mention: since I'm looking at this in the context of diffusion-limited aggregation, my points are not a cloud but are all connected, hopefully in an interesting-looking structure. Here's what I've been working on - I started with some code posted a while ago on the J wiki: http://www.jsoftware.com/jwiki/Studio/Gallery/DLA; did some work revising it here - http://www.jsoftware.com/jwiki/DevonMcCormick/DLA00 - then continued in this area of the wiki at "DLA0Runs" and "DLA01". We also talked about this at NYCJUG this month: NYCJUG/2009-10-13#DLA.3AInitialJImplementation. I'm looking to improve this algorithm by concentrating only on the periphery of the cluster, hence my attempts to programmatically define the perimeter. Regards, Devon On Sat, Oct 24, 2009 at 11:43 PM, Brian Schott <[email protected]>wrote: > If you want to account for a more elliptical than circular shape of > the cloud of points, or if you want to deal more with higher > dimensional points, consider Mahalanobis distance . > > http://en.wikipedia.org/wiki/Mahalanobis_distance > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > -- Devon McCormick, CFA ^me^ at acm. org is my preferred e-mail ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
