-------- Original Message --------
Subject: Re: Non-independent groups in DA
Date: Fri, 31 Oct 2008 11:04:18 -0700 (PDT)
From: F. James Rohlf <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
To: Morphmet <[email protected]>
If I understand the question, there is no problem. There is no
assumption of independence of groups. They simply have to be known a
priori. Assumptions to worry about are independence of sampling of
specimens and homogeneity of covariance matrices.
------Original Message------
From: morphmet
To: Morphmet
ReplyTo: Morphmet
Sent: Oct 31, 2008 8:54 AM
Subject: Non-independent groups in DA
-------- Original Message --------
Subject: Non-independent groups in DA
Date: Thu, 30 Oct 2008 17:41:12 -0700 (PDT)
From: Paul Sanfilippo <[EMAIL PROTECTED]>
To: [email protected]
Dear Group,
I have a sample of optic discs that I have used as the basis for
creating 2 other groups with slightly different overall shapes
compared to the originals. The purpose of this was to run some
analyses with different techniques (eg EFA, SSL analysis, Eigenshape
analysis) and then perform discriminant analysis to see how well the
groups are classified for each technique. Obviously the 3 sample
groups are not independent and my question is does this non-
independence affect the validity of the classification table (% of
correctly classified specimens according to groups) that the DA
calculates? I used a cross-validation (leave one out) for this.
Thank you,
Paul Sanfilippo
Uni Melbourne,
Australia
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
--------
Sent remotely by F. James Rohlf
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
