Date:         Wed, 24 May 2006 10:25:02 -0400
Reply-To:     Peter Flom <flom@NDRI.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <flom@NDRI.ORG>
Subject:      Re: Problems using PROC GLM to fit a covariate model
Comments: To: Chris.Howden@CBR.COM.AU
Content-Type: text/plain; charset=US-ASCII

Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
http://cduhr.ndri.org
www.peterflom.com
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)


>>> Chris Howden <Chris.Howden@CBR.COM.AU> 05/24/06 3:43 AM >>> wrote
<<<
SAS is having some computational problems when fitting a covariate model
and I'm a little unsure as to wether it's a algorthmic problem or a
statistical problem.

It's a simple covariate model with 1 covariate and 1 factor with 7
levels (2 reps in each level).

The covariate varies between levels but NOT within levels...and this
appears to be the problem. SAS cannot fit this model. However if I
introduce within level variation of the covariate then SAS can fit the
model.

This confuses me since I can think of no theoretical reason why a
covariate needs to have within level variation for the model to be
valid....hence I'm concluding its an algorithimic problem with SAS.
>>>

No problem with SAS, big problem with your model

Your two independent variables are completely colinear.
There is no way to evaluate this model, you must either:

1) Eliminate either the covariate or the factor - this won't hurt your
model,
since the factor is equivalent to the covariate

2) Get more data, where the covariate does vary within the factor.  Even
then,
a simple GLM is unlikely to be right, since, unless you get a HUGE
amount of new
data, there will still be substantial colinearity.


HTH

Peter