Date: Fri, 30 Nov 2007 17:52:05 -0800
Reply-To: Ching <KCWong5@GMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Ching <KCWong5@GMAIL.COM>
Organization: http://groups.google.com
Subject: Re: logistic output.
Content-Type: text/plain; charset=ISO-8859-1
On Nov 30, 1:53 am, raghupath...@GENPACT.COM ("N, Raghupathi") wrote:
> Hi all,
>
> Could you please check?
> 1. Whether you are satisfied Convergence criteria if not the parameter
> Estimates are not found to be stable?)
> 2. It occurs mainly due to collinearity effects.
>
> Provide the completeoutputoflogisticregression so that it make clear
> to all of us.
>
> raghup
>
>
>
> -----Original Message-----
> From: SAS(r) Discussion [mailto:SA...@LISTSERV.UGA.EDU] On Behalf Of
>
> Ching
> Sent: Thursday, November 29, 2007 4:57 AM
> To: SA...@LISTSERV.UGA.EDU
> Subject: Re:logisticoutput.
>
> On Nov 29, 4:08 am, sonik son <sonik...@gmail.com> wrote:
> > On Nov 28, 5:06 am, peterflomconsult...@mindspring.com (Peter Flom)
> > wrote:
>
> > > >> Ching <KCWo...@GMAIL.COM> wrote
> > > >> >could anyone please tell me if it is still right to report the
> OR
> > > >> >below of Grp2 v 1 and 3 v 1 after adjusting for gender and
> smoking
> > > >> >status given below is theoutput?
> > > >> >if not, what should I do? exlude gender and smoker?
> > > >> >thanks heaps.
> > > >> >have a wonderful day.
>
> > > >> > TheLOGISTICProcedure
> > > >> > Odds Ratio Estimates
>
> > > >> > Point 95% Wald
> > > >> >Effect Estimate Confidence Limits
>
> > > >> >age_var 1.047 0.950 1.155
> > > >> >male >999.999 <0.001 >999.999
> > > >> >smoker 2 vs 1 0.245 0.024 2.467
> > > >> >smoker 3 vs 1 <0.001 <0.001 >999.999
> > > >> >Grp 2 vs 1 6.842 0.687 68.146
> > > >> >Grp 3 vs 1 <0.001 <0.001 >999.999
>
> > > I replied
>
> > > >> Something is clearly wrong.
>
> > > >> How many men and women were in the study? How many in Grp 1, 2,
> and 3?
> > > >> How many in Smoker 1,2,3?
>
> > > >> How were each of these coded? (ie. was male = 1, female = 0 or
> what?)
>
> > > Ching replied
>
> > > >if male then male=1 else male=0.
> > > >the number of male=290 and female is 10
> > > >the number is 70, 90,25 for smoker=1, 2 and 3 respectively.
> > > >the number is 180, 90 and 30 for grp=1, 2 and 3 respectively.
> > > >thanks!
>
> > > My new reply
>
> > > The problem is too few women. You can either
> > > a) Gather data on more women (if possible)
> > > b) Do the analysis only on the 290 men
> > > or
> > > c) Do the analysis on all 300, but ignore sex. This is probably not
> a good idea, unless you are certain that sex is irrelevant for your
> study
>
> > > Peter- Hide quoted text -
>
> > > - Show quoted text -
>
> > I agree with Peter that the number of women in the sample is too low.
> > Central Limit Theorum recommends at least 30 observations for the
> > assumptions of normality to hold for "good" statoutput.- Hide quoted
> text -
>
> > - Show quoted text -
>
> thanks everyone =)- Hide quoted text -
>
> - Show quoted text -
Hi there.
SAS says
WARNING: There is possibly a quasi-complete separation of data points.
The maximum likelihood
estimate may not exist.
WARNING: The LOGISTIC procedure continues in spite of the above
warning. Results shown are
based on the last maximum likelihood iteration. Validity of
the model fit is
questionable
thanks.
have a great weekend~
| 
