Date: Fri, 3 Jan 1997 09:28:16 -0800
Reply-To: Simo Virtanen <svir@OCCUPHEALTH.FI>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Simo Virtanen <svir@OCCUPHEALTH.FI>
Organization: Finnish Institute of Occupational Health
Subject: Re: Odds Ratios
Content-Type: text/plain; charset=us-ascii
Paul Robertson wrote:
----- cut -----
> sepsis and trauma patients. The odds ratio I obtained
> was about 4.
> How does this square with the fact that the mortailty
> for the sepsis patients was 53% and for the trauma
> patients it was about 22%? Is the problem the intercept?
> When I ran the logistic regression without an intercept
> the odds ratio obtained was around 1.3. -- dividing 53 by
> 22 would indicate about twice the risk of death. right?
No, an odds ratio of about 4 seems right to me.
The **odds** for the first group would be 53/47=1.127,
and odds for the second group would be 22/78=0.282.
Then **odds ratio** would be 1.127/0.282=3.996.
Note that the odds are defined as chances of
being in one category as opposed being in others.
Simo V. Virtanen, Ph.D.
Finnish Institute of Occupational Health