Date: Tue, 6 Dec 2005 08:13:21 -0500
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
Subject: Re: PROC LOGISTIC: Odds ratios with interactions
Content-Type: text/plain; charset=us-ascii
On Mon, 5 Dec 2005 19:42:05 -0800, davidlcassell@MSN.COM (David L
>>IMHO at least, interaction effects in logistic models are intepretable
>>and make sense. There is a complete monograph named "Interaction
>>Effects in Logistic Regression" that addresses just that and nothing
>>else: interaction effects in logistic regression models. Tha author is
>>James Jaccard of State University of New York, Albany, NY in the
>>Series: Quantitative Applications in the Social Sciences. Volume # 135.
>>If it is not in your local/company/campus library, you may get it for
>>16US depending on where you buy it. ISBN: 0-7619-2207-5. See for
>>For a milder treatment of interactions in logistic regression, you
>>might want to see the classical "Categorical Data Analysis" by A.
>>Agresti. There is one example in Chapter 7, I think, that fits and
>>interprets a logistic model with interactions.
>I agree completely. Interactions in logistic regressions are just as
>interactions in linear regressions.
>But that's not quite the problem of the original poster. He was modeling
>interaction terms, and then expecting to get meaningful odds ratios for
>the main effects. My point was that if you have significant interaction
>then you don't have meaningful odds ratios for the main effects that have
Well it depends on the size of the effect doesn't it? In a large
dataset even a tiny interaction can be statistically significant, in
which case interpretation of main effects might still make good sense.
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.