A PCA (if you use all dimensions) preserves distances. Thus, distances based on 
the initial data and PCA scores using all dimensions should be identical. Note: 
if the PCA is based on a correlation matrix rather than a covariance matrix 
then you have to use standardized data to get equality. Mahalanobis distances 
are different as are Procrustes distances. 


F. James Rohlf                                    
Distinguished Professor, Emeritus
Depts: Anthropology and Ecology & Evolution
Stony Brook University
On 12/15/2020 9:47:38 AM, Bryan McLean <[email protected]> wrote:
Dear morphometricians,

I have a basic question that some quick literature searches havent answered 
yet, but which I believe will be answered quickly from the collective knowledge 
here. 

I wish to compare pairwise distances among specimens in PCA space derived from 
a) a suite of linear measurements and b) a 3D landmark-based approach. The goal 
is simply to ask whether one of the two approaches to describing shape provides 
"higher resolution" (i.e., more discriminating power). However, I dont think 
that raw (Euclidean or Mahalanobis) distances from the separate PCAs are 
comparable, since they explain different amounts of total variation. 

Is there a way to scale the among-specimen distances from each so they are 
comparable between the PCAs? 
OR - is there a totally different (eg non-PCA-based) way of approaching this 
simple question?

Thanks in advance,
Bryan
--
You received this message because you are subscribed to the Google Groups 
"Morphmet" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected] 
[mailto:[email protected]].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/morphmet2/CAFXGjdoCoKZOUZ-iFPRM4UuKxByzCt%2BKcksxs%2Bwew0tfHxK70g%40mail.gmail.com
 
[https://groups.google.com/d/msgid/morphmet2/CAFXGjdoCoKZOUZ-iFPRM4UuKxByzCt%2BKcksxs%2Bwew0tfHxK70g%40mail.gmail.com?utm_medium=email&amp;utm_source=footer].

-- 
You received this message because you are subscribed to the Google Groups 
"Morphmet" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/morphmet2/Mailbird-6abbbbf7-60c4-4c0f-a3eb-a49f1e20999e%40stonybrook.edu.

Reply via email to