Re: Canonical correlation for clustering?
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Thu, 24 Aug 2006 16:07:18 -0400
On 24 Aug 2006 00:51:53 -0700, sanketrepo@xxxxxxxxx wrote:
I was just wondering if the Cannonical correlation has been applied in
any industry or research for clustering. essentially the idea would be
to cluster points(clusters) together based on their canonical
correlation coefficients. Seems to have some potential but could not
find any relevant use in applications. Please give your comments and
suggestions.
The mathematics of canonical correlation is what underlies
discriminant function and its generalization as MANOVA,
as well as 'multivariate multiple regression.'
Correspondence analysis is also built on canonical correlation,
and its results are often expressed as plots that look like
useful clusters, as defined separately by the two sets of variables.
You can Google on correspondence-analysis. Oklahoma State
University has a site that includes other 'ordination' methods, too.
They show a variety of techniques as used in ecological studies,
starting at
http://ordination.okstate.edu/
--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.
- References:
- Canonical correlation for clustering?
- From: sanketrepo
- Canonical correlation for clustering?
- Prev by Date: analyzing leadership "priorities"
- Next by Date: Re: Reporting effect sizes for non-significant results
- Previous by thread: Re: Canonical correlation for clustering?
- Next by thread: Splitting an SPSS file to create two matched groups
- Index(es):
Relevant Pages
|
