Re: Numerical approach to structure asymmetry

From: Ray Koopman (koopman_at_sfu.ca)
Date: 03/25/05 

Date: 25 Mar 2005 12:36:30 -0800

Krzysztof Nowak wrote:
> I'm going to study planar 2D structure built with 10-15 points.
> How could some find anisotrophy of this distribution? It would be
> nice to measure size of anisotrophy and consider weight of
> particular points.
>
> My first approach is to calculate the most appropriate axis with
> linear regression, than somehow calculate 'length' and 'width' of
> structure. Assymetry could be (length - width) / (length + width).
>
> It should work, but I was quite shure that it is a common problem
> and I could find some smart solution (now, after couple of hours
> of browsing I'm not so shure anymore). I need to calculate huge
> amount of data and I would really appreciate something more clever.
>
> Any ideas?

Shift the origin to the centroid. The singular values of the centered
matrix are your 'length' and 'width'.