----- Forwarded message from Øyvind Hammer -----
Date: Thu, 15 Nov 2012 15:57:44 -0500
From: Øyvind Hammer
Reply-To: Øyvind Hammer
Subject: Re: UPGMA: Euclidean and Procrustes distances
To: [email protected]
> 1 - Are UPGMA analyses using Euclidean distances and Procrustes
> distances equivalent?
Procrustes distances are approximately (opening-can-of-worms alert) the
same as Euclidean distances between Procrustes-fitted landmark
configurations. In Past you can Procrustes-fit your data first, and then
use UPGMA with the Euclidean measure.
> 2 - Is there any package in which I could simply enter the Procrustes
> distances and get a UPGMA tree with bootstrap values?
I believe that would be impossible, because you must bootstrap the
original data, not the distance matrix. A new distance matrix must be
computed from the resampled variables for each bootstrap replicate, as
Past does.
I will leave to others to debate whether bootstrapping of landmarks for
UPGMA is a good idea. Also, Past will resample the data columns
individually, without regard for whether they are X or Y coordinates of
landmarks, I'm not sure what that implies.
Øyvind Hammer
(Man of the Past)
----- End forwarded message -----
