RE: PCA and elliptical confidence

[morphmet_moderator]

----- Forwarded message from "F. James Rohlf" -----

Date: Fri, 30 Nov 2012 12:59:49 -0500
From: "F. James Rohlf"
Reply-To: ro...@life.bio.sunysb.edu
Subject: RE: PCA and elliptical confidence
To: morphmet@morphometrics.org

A confidence region shows the uncertainty for the estimate of the mean. It does not imply that individuals from the groups can be distinguished. With very large sample sizes the confidence regions will be very small but even if they do not overlap the scatter of individual observations may broadly overlap.  Perhaps you should compute equal-frequency ellipses – their purpose is to show the range of dispersion of individuals. The different sizes of these will give you information about differences in sample variances.

 

----------------------

F. James Rohlf, John S. Toll Professor, Stony Brook University

The much revised 4^th editions of Biometry and Statistical Tables are now available:

http://www.whfreeman.com/Catalog/product/biometry-fourthedition-sokal

http://www.whfreeman.com/Catalog/product/statisticaltables-fourthedition-rohlf

P Please consider the environment before printing this email

 

 

From: morphmet_modera...@morphometrics.org [mailto:morphmet_modera...@morphometrics.org]
Sent: Friday, November 30, 2012 1:33 AM
To: morphmet@morphometrics.org
Subject: PCA and elliptical confidence

 

----- Forwarded message from dan dkin -----

Date: Thu, 29 Nov 2012 06:19:37 -0500
From: dan dkin
Reply-To: dan dkin
Subject: PCA and elliptical confidence
To: morphmet@morphometrics.org

 

Hi,

When I run a PCA and applying elliptical confidence I can distinguish groups. The ellipses I get are in different radiuses and almost do not overlap. Therefore, except the difference of the groups does the ellipse shape tell something (e.g., variance in the groups, etc).

 

Any explanation, paper or etc will be welcome.

 

Thank

 

----- End forwarded message -----

 

 

----- End forwarded message -----