Date:         Sun, 18 Aug 1996 00:55:38 -0400
Reply-To:     Ellen Hertz <eshertz@ACCESS.DIGEX.NET>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Ellen Hertz <eshertz@ACCESS.DIGEX.NET>
Organization: Express Access Online Communications, USA
Subject:      Re: SAS logistic regression

jacqui@coredcs.com wrote:
>
> I have adapted the match_11 logistic regression program
> to my data set as outlined on pp. 112-113 in "Logistic
> Regression:  Examples using the SAS sytem".  This program
> requires a "noint" option in the model statement.  However,
> I need an intercept to complete the model equation and
> calculate probabilities.  I have tried several options such
> as excluding the noint option or substitution of a number of
> other options which only caused the program to fail.  Can
> anyone provide an answer to this dilemma?
>
>
>
> -----------------------------------------------------------------------
> This article was posted to Usenet via the Posting Service at Deja News:
>                    http://www.dejanews.com/
> Please report abuse of this service to: postnews@dejanews.com
 
 
Match_11 conditional logistic regression does not calculate
unconditional probabilities of a positive response;
the coefficients represent the effect of the
variables on the log odds of being the case (as opposed to the
control) given that there is one of each in the stratum.
 
If you want to estimate probabilities without the noint option
it might be more appropriate to use ordinary logistic regression,
making each subject an observation instead of the pair. The
stratifying variables on which the case/control pairs were matched
should be included in the model in that case.