Re: eigenshape error

[morphmet]

In my current standard eigenshape package there is an option that 
allows you to specify the error tolerance you allow in representing 
the shape. In most cases the number of points you need to accurately 
represent a complex curve is surprisingly small. This issue is part 
of what led me to develop the extended eigenshape method and the 
tolerance sampling routine is integral to that approach. As for error 
estimation in the ordinations, that can be approached in a variety of 
ways depending on what type of error you are interested in. In this 
context you should remember that eigenshape is simply a singular 
value decomposition of a set of shape functions. Error estimation 
inherent to the sample of shapes can be estimated in the same way it 
is for any data set for which SVD is applicable. Or, if you are just 
using the method as a shape-transformation technique, error can be 
ignored entirely.

Norm MacLeod




>All,
>
>Forgive my ignorance if this is a stupid question, but does anyone know
>how to assess the error in eigenshapes??  I assume the accuracy with
>which the curve is captured depends on the number of points and is
>user-defined, and the precision could be constrained by making repeat
>measures, but how does that translate into, for example, placing error
>bars on an ordination, if that even makes sense?
>
>Thanks
>
>David
>
>................................................
>David Jones
>Ph.D. researcher, Palaeobiology Research Group
>Department of Geology
>University of Leicester
>University Road
>Leicester  LE1 7RH
>Tel.: +44(0)116 252 3912 Ext. 3632
>==
>Replies will be sent to list.
>For more information see
http://life.bio.sunysb.edu/morph/morphmet.html.


-- 


___________________________________________________________________

Dr. Norman MacLeod
Keeper of Palaeontology
The Natural History Museum,
Cromwell Road, London, SW7 5BD


(0)20-7942-5204 (Office)
(0)20-7942-5546 (Fax)
http://www.nhm.ac.uk/palaeontology/a&ss/nm/nm.html (Web Page)

___________________________________________________________________
==
Replies will be sent to list.
For more information see http://life.bio.sunysb.edu/morph/morphmet.html.