Date:         Mon, 12 Oct 2009 07:23:54 -0400
Reply-To:     Muthia Kachirayan <muthia.kachirayan@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Muthia Kachirayan <muthia.kachirayan@GMAIL.COM>
Subject:      Re: Relative risk and odds ratio, OR significant and above 1,
              RR not
In-Reply-To:  <28880866.1255344114615.JavaMail.root@mswamui-cedar.atl.sa.earthlink.net>
Content-Type: text/plain; charset=ISO-8859-1

On Mon, Oct 12, 2009 at 6:41 AM, Peter Flom <
peterflomconsulting@mindspring.com> wrote:

> Thomas Fröjd <thomas.frojd@NEURO.UU.SE> wrote
>  >I am performing a logistic regression for Odds Ratio as result and a
> >Poisson regression with log link for RR, both with proc genmod. The
> >design is a cohort study with two timepoints sampled so far.
> >
> >A thing that confounds me is that one of the covariates has a pretty
> >strong and significant effect when regressing for Odds Ratio but then
> >points to a negative but unsignificant when regressing for Relative Risk.
> >
> >Anyone can explain why to me?
>
> The positive vs. negative change might be due to how SAS codes things, you
> might need the DESCEDING option in LOGISITC (look at the output to see how
> SAS is coding things).
>
> But what's your DV?  Is it a count (therefore POISSON might be appropriate)
> or dichotomous (LOGISTIC) or ordinal (also LOGISTIC) or what?
>
> And, more generally, since RR and OR ask somewhat different questions, they
> can give different answers.
>
> Without more context (and maybe code) it's hard to say what's going on.
>
> Peter
>
> Peter L. Flom, PhD
> Statistical Consultant
> Website: www DOT peterflomconsulting DOT com
> Writing; http://www.associatedcontent.com/user/582880/peter_flom.html
> Twitter:   @peterflom
>

In Logistic regression the DV is ln(p/q) ie the log odds. The use of words -
regressing Odds Ratio in " .......significant effect when regressing for
Odds Ratio ....", seem to go against log odds as Odds Ratio is defined as
(p1/q1) / (p2/q2).

Further, RR uses a ratio like p1/p2, where p1 is rate in Exposed Group and
p2 in Unexposed Group. How p1/p2 is transformed to a DV to use  LR?  I have
not used RR in this context or I may be wrong. The OP may tell how was this
done.

Muthia Kachirayan