Date: Thu, 24 Aug 2006 01:01:23 -0700
Reply-To: Albert-jan Roskam <fomcl@YAHOO.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Albert-jan Roskam <fomcl@YAHOO.COM>
Subject: Re: catmod and empty cell
Content-Type: text/plain; charset=iso-8859-1
I agree with David. Start with a good research
question, and determine what an empty cell really
means. If the variable is, for example, amount of
smoking per day, zero would mean 0 cigs/day, whereas
null (empty cell) would mean 'data not known', for
example because the respondent didn't fill in the
question, or because of a scanning error, or whatever.
Thinketh before thy recodeth ;-)
My question: what proc could I use for multinomial
hierarchical logistic regression, considering that I
use SAS v8.02. Proc logistic is only possible from v9,
right? So should I use proc genmod or proc catmod, or
yet something else?
Thank you in advance!
--- David L Cassell <davidlcassell@MSN.COM> wrote:
> adel_tangi@YAHOO.FR wrote:
> > I have two categorical variables with 4
> categories each, I want to
> >analyse the association between the two variables,
> I think about the
> >multinomial logistic regression, but, I have one
> cell with 0 count (empty
> > How to model the association correctly?
> > Thanks
> > Adel
> Do you really need to *model* the association, or do
> you just want to test
> whether the association is significant? In the
> second case, I would
> using PROC FREQ with the EXACT statement. If it
> takes forever to run,
> I would re-run it with the EXACT statement and the
> Monte Carlo option.
> If you really want to model that association, then
> you next have to
> decide whether that zero is structural: is it
> supposed to be there based
> on the theory of the variables? The answer to that
> will determine where
> you go next.
> And also, are you sure you want ot use PROC CATMOD?
> It's effective,
> but there are lots of newer procs you might be able
> to use here.
> David L. Cassell
> mathematical statistician
> Design Pathways
> 3115 NW Norwood Pl.
> Corvallis OR 97330
> Get the new Windows Live Messenger!
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around