Date:         Sat, 16 Nov 1996 20:56:51 -0500
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: DIGEX
Subject:      Re: plot logistic regression output

Gema wrote:
>
> Can anyone help me out in plotting logistic regression results to test
> for linearity?
>
> Thanks--
>
> GemaGema-
 
The results will return the logit of the probablity of a positive
outcome as a linear function of the independent variable(s). If there is
only one independent variable you might be interested in plotting
the logit of the proportion of actual positive outcomes at each value
(or small interval of values) of the independent variable to see if it
looks linear.
 
If your purpose is to test the fit of the model you might
look at the measure of concordance in the output or you could compare
the likelihood to that of other candidate models that are functions
of the same independent variables. There is also a test described in
Hosmer and Lemeshow that partitions the data into deciles of
(predicted) risk and uses a chi square that is based on comparing
actual number of positive responses in each decile with the
predicted number.
 
Hope this helps.
 
Ellen