Thanks again to everyone for your helpful answers. I think I will now be
able to fix my problem to  the possible extent.

Louis

> Date: Fri, 9 Nov 2007 11:28:40 +0100
> From: [EMAIL PROTECTED]
> To: [email protected]
> Subject: Re: PCA clusters
>
> ... as regarding the "recognition" of groups in a
> continuum without prior group assignment, one method
> not mentioned so far is mixture modelling, which is
> simply attempting to fit the data to a mixture of
> e.g. gaussians. Criteriae such as the Akaike Information
> Criterion may be used to get some idea about a "reasonable"
> number of groups, although a proper significance test
> is not really available.
>
> And then you have all those old clustering methods,
> including more or less desperate attempts to calculate
> some kind of significance (beware).
>
> If the points are assigned to groups from independent
> information, there are the standard statistical tests.
> Since you are primarily interested in PC1, a one-way
> ANOVA might be sufficient, although I think a MANOVA on
> all PCs would be better.
>
> All these methods are available in e.g. R or PAST.
>
>
>
> Oyvind Hammer
> Natural History Museum
> University of Oslo
>
>
>
> --
> Replies will be sent to the list.
> For more information visit http://www.morphometrics.org
>

------------------------------------------------------------------------
Machen Sie Ihre E-Mail-Kontakte zu Messenger-Freunden! Einfacher
Adressimport!
<http://redirect.gimas.net/?cat=hmtl&n=M1007AI&d=http://messenger.live.de/ersteschritte_adressimport.html>

-- 
Replies will be sent to the list.
For more information visit http://www.morphometrics.org

Reply via email to