Date:         Tue, 6 Dec 2005 10:59:57 -0800
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: PROC LOGISTIC: Odds ratios with interactions
In-Reply-To:  <200512061330.jB6Bk3HF020379@mailgw.cc.uga.edu>
Content-Type: text/plain; format=flowed

Jay replied (to my comment):
>Well it depends on the size of the effect doesn't it?

Yes.  I should have been clearer.  If the interactions are minute but
statistically significant, then their effect on the main effects may be so
small that odds ratios of A computed at each level of B could be
essentially identical.

>                                                        In a large
>dataset even a tiny interaction can be statistically significant, in
>which case interpretation of main effects might still make good sense.

Yes.  Unfortunately, with large enough data sets, one can get *everything*
to be statistically significant.  Even when the paramter estimates are so
small
that they have no scientific or business-case meaning.

>I've recently completed logistic regression analyses of tens of
>thousands of medical records in which virtually every minuscule 3- &
>4-way interaction was significant at the .001 level.

Ugh.  You have my deepest sympathies.  :-)  :-)

David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
Don’t just search. Find. Check out the new MSN Search!
http://search.msn.click-url.com/go/onm00200636ave/direct/01/