Date:         Thu, 4 Nov 2010 19:38:44 -0400
Reply-To:     Jordan H <jihool3670@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jordan H <jihool3670@GMAIL.COM>
Subject:      Re: huge (>999.99) odds ratios: cause?
In-Reply-To:  <C4A8805C2D21D643B4DA7DF789C4FEB10FCC967D@VANCRMSGA2.vha.med.va.gov>
Content-Type: text/plain; charset=ISO-8859-1

Thank you, Peter, Dean, and Zack (who emailed me privately), for your
insightful responses.

I believe I checked the cross tabs prior to sending out my original email
but I shall check again when I have the chance.  The vast majority of the
variables are binary so it's likely the cross tab will only be a 4x4 matrix.
 If it turns out that the odds ratios are being produced by near 0 cells, is
my only option to drop the variable, since collapsing levels of the variable
isn't a possibility with binary variables?

As per usual, thanks for the consideration.  This listserv is truly a
fantastic resource.

Jordan



On Tue, Nov 2, 2010 at 2:56 PM, Bross, Dean S <dean.bross@va.gov> wrote:

> An odds ratio of 999.99 usually means to me that everybody died in that
> group.
> So the odds in that group are infinity.
>
>
> -----Original Message-----
> From: owner-sas-l@listserv.uga.edu [mailto:owner-sas-l@listserv.uga.edu]
> On Behalf Of Jordan H
> Sent: Tuesday, November 02, 2010 2:17 PM
> To: sas-l@listserv.uga.edu
> Subject: huge (>999.99) odds ratios: cause?
>
> Hello, all.
>
> First, a little background.  I've been asked to help with a project in
> which
> the goal to develop a model that predicts high cost pharmacy
> expenditures
> based on a variety of variables, such co-morbidities, demographics, etc.
> To
> do this, a multivariate regression model was used.  My client is also
> interested in trying to model poor prediction within the multiple
> regression
> model.  To do this, they saved the residuals from  PROC REG, made an
> indicator variable for those observations with residuals greater than
> 1.75,
> and ran a PROC LOGISTIC with the new indicator variable as the response
> variable and the original independent variables, plus additional cost
> variables, as predictors.
>
> The model converges and most coefficients/odds ratios look reasonable
> but
> some appear to be errors (odds ratios of >999.99, confidence intervals
> (<0.001 - >999.99).  We've checked things like multicollinearity but
> that
> doesn't seem to be an issue.
>
> Does anyone have an idea as to what could be going on?
>
> Thank you for your consideration!
> Jordan
>