Date:         Thu, 1 Sep 2005 09:59:33 -0300
Reply-To:     Hector Maletta <hmaletta@fibertel.com.ar>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Hector Maletta <hmaletta@fibertel.com.ar>
Subject:      Re: Decomposing interaction term in Cox Regression
Comments: To: "David M. Fresco" <fresco@kent.edu>
In-Reply-To:  <D5D0B9C4-BAC6-4274-88E1-E7B4B294A4AA@kent.edu>
Content-Type: text/plain; charset="us-ascii"

Just do the same. Plot the hazard curve for various combinations of levels
of A an B (high-high, high-low, low-high, low-low, for instance). You can do
this in various way. One is generating (with COMPUTE) the value of the
function for each subject at various given levels of A and B, plus actual
values of the other covariates, then plot the results.

Hector

> -----Original Message-----
> From: David M. Fresco [mailto:fresco@kent.edu]
> Sent: Thursday, September 01, 2005 9:53 AM
> To: Hector Maletta
> Subject: Re: Decomposing interaction term in Cox Regression
>
>
> On 01 Sep 2005, at 8:47, Hector Maletta wrote:
>
> > David,
> > I don't understand what your difficulty is, since you say
> you have a
> > good grasp of interactions in regression. The idea is basically the
> > same. In this case, you have found that the hazard rate for relapse
> > increases with A (at constant B), with B (and constant A) and
> > furthermore these two factors reinforce or potentiate each other as
> > indicated by the significant interaction of both. This increase in
> > hazard for subjects with higher AB product, of course, is
> supposed to
> > be proportional along time, i.e. to affect all patients at
> all times.
> > Suppose A=sex and B=smoker, you found that one sex has more
> risk than
> > the other, smoking also increases the risk, and furthermore, the
> > effect of smoking is dependent on sex and vice versa.
> >
> > Hector
> >
>
> Thanks for the reply.  I should have added that A and B are
> both continuous measures.  The way I handle such interaction
> terms in OLS is to solve the full model at various levels of
> the predictors in a manner described in Cohen and Cohen.
> Although I understand OLS pretty well, I am a bit out of my
> element in interpretting and reporting the hazard values that
> would emerge from some decomposition of my interaction term.
>
> Any further suggestions?
>
> ------
> David M. Fresco, Ph.D.               fresco@kent.edu
> Department of Psychology        Voice:  (330) 672-4049
> Kent State University                   Fax:    (330) 672-3786
> 226 Kent Hall Annex
> P.O. Box 5190                               http://dept.kent.edu/
> psychology/
> Kent, OH  44242-0001                http://www.personal.kent.edu/
> ~dfresco/
>
> ListOwner of Helplessness http://psych.upenn.edu/~fresco/
> helplessness.html Listowner of Mindfulness
> http://listserv.kent.edu/archives/
> mindfulness.html
>
>
>
> __________ Informacisn de NOD32 1.1207 (20050831) __________
>
> Este mensaje ha sido analizado con  NOD32 Antivirus System
> http://www.nod32.com
>
>