Date:         Wed, 29 Sep 2004 09:45:44 -0700
Reply-To:     Dale McLerran <stringplayer_2@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Dale McLerran <stringplayer_2@YAHOO.COM>
Subject:      Re: using Logistic regression to obtain Relative Risks
In-Reply-To:  <6.1.2.0.0.20040929172844.0537adf8@10.0.10.66>
Content-Type: text/plain; charset=us-ascii

--- Henric Nilsson <henric.nilsson@STATISTICON.SE> wrote:

> Hi,
>
> At 15:32 2004-09-28 -0400, you wrote:
>
> >I'm reviewing a manuscript that includes the following paragraph:
> >Quote on:
> >We estimated relative risks, confidence intervals and p-values using
> a form of
> >the generalized linear model, substituting the log of the risk ratio
> as the
> >link
> >function and assuming the residuals from the maximum likelihood
> fitting
> >procedure
> >follow a Poisson distribution.  This method produces regression
> >coefficients that,
> >after exponentiating, can be interpreted as relative risks.
> >
> >The description of what these authors did is not clear to me, nor is
> >it necessarily the case that they used SAS.
>
> No, the description is definitely lacking. And not entirely correct,
> it
> seems to me.
>

I would agree that the description is not clear nor is it entirely
correct.  Specifically, it is not residuals that are assumed to
follow a Poisson distribution.  It is the response, conditional on
X'beta (or, for a log link, exp(X'beta)), that is assumed to be
distributed as a Poisson variate.  That should be pointed out to
the authors.

Dale


=====
---------------------------------------
Dale McLerran
Fred Hutchinson Cancer Research Center
mailto: dmclerra@NO_SPAMfhcrc.org
Ph:  (206) 667-2926
Fax: (206) 667-5977
---------------------------------------



_______________________________
Do you Yahoo!?
Declare Yourself - Register online to vote today!
http://vote.yahoo.com