Date:         Fri, 10 Oct 2003 09:35:18 -0700
Reply-To:     Allison Ellman <allison@QUONET.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Allison Ellman <allison@QUONET.com>
Subject:      Re: test for this IV, these DVs and covars
Content-Type: text/plain; charset="us-ascii"

Hi,

Are you saying that the covariates are categorical?  I think the proper
way to handle them is to make dummy variables (a set of dichotomous
variables coded 0 for all categories except one) For instance, for race,
a dummy variable Caucasian will equal 0 for everyone except Caucasians
for which it will equal 1.  africanamerican will equal 0 for everyone
except African Americans. Etc.  Then, all but one of the dummy variables
are included in the model (the one left out is the reference group.

Hope this helps.

Allison

. . .
Allison Ellman

-----Original Message-----
From: Nico Peruzzi, Ph.D. [mailto:nperuzzi@yahoo.com]
Sent: Friday, October 10, 2003 9:20 AM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: test for this IV, these DVs and covars

Hi Listers,

I've got myself a little confused and am hoping someone can
guide me back onto track.

My main factor (drug group) has three levels.
I have 5 numeric outcomes measures of interest.
I have a number of covariates (both numeric and
categorical).

I want to examine the effect of my main factor on the
outcome variables, while removing the influence of the
covariates.

Seems simple enough, but I find that the GLM models seem to
specify using only numeric covariates.  I then thought
about using DISCRIM, but got confused as how to step in the
outcomes measures versus the covariates, and well, here I
am...

Any guidance would be greatly appreciated.

Thanks,  Nico

__________________________________
Do you Yahoo!?
The New Yahoo! Shopping - with improved product search
http://shopping.yahoo.com