Date:         Sat, 3 Sep 2005 02:33:59 -0700
Reply-To:     Mustela <mustela@MCE.HGGM.ES>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Mustela <mustela@MCE.HGGM.ES>
Organization: http://groups.google.com
Subject:      enough degrees of freedom in a repeated design
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset="iso-8859-1"

Hi,

My data consists of five subjects that have been measured using five
different machines. I have several variables measured with the data
from each machine. This is a test intended to verify that I can mix
data from different machines without introducing too much error.

To test the significance of the machine factor in each of my variables,
I´m using proc GLM and machine as a repeated measure.


proc glm DATA = repeats ;
 model MACH_1 MACH_2 MACH_3 MACH_4 MACH_5 =  ;
 REPEATED machine 5/  SUMMARY ;

My question whether I have enough degrees of freedom or not. Five
subjects, five machines, total n=25. I´m suspicious of the validity of
the assumptions because in most of the variables results are highly
significant, but looking at the plots there´s little agreement between
the differences that you´d expect and the significance of the tests.

Many thanks in advance!

Santiago