Date:         Wed, 10 Jan 2007 19:12:37 -0500
Reply-To:     Wensui Liu <liuwensui@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Wensui Liu <liuwensui@GMAIL.COM>
Subject:      Re: Contrast Statement in PROC LOGISTIC
Comments: To: "Earley, Joseph" <JEarley@lmu.edu>
In-Reply-To:  <E377388CEF06ED40AEA2C1CC5A93687102BBDDD6@Aries2.lmumain.edu>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

joseph,

do you want to do a likelihood ratio test between 2 models.

if yes, then it should be very easy to do using loglikehood from 2
models. the ratio should follow chi-square with df equal to the the
difference of model df between 2 models.

On 1/10/07, Earley, Joseph <JEarley@lmu.edu> wrote:
> Is anyone familiar with the coding for the CONTRAST option in PROC
> LOGISTIC?
>
>
>
> We are estimating Y = f( X1, X2, X3)  then adding X4 which is a 0/1
> dummy.
>
>
>
> The task is to make sure that we have the correct coding for  the
> CONTRAST statement to perform the LR test for
>
>
>
> The two models:
>
>
>
>  y = f( x1, x2 and x3)  versus   Y = f( x1, x2, x3 and x4).
>
>
>
> Thanks for any code example.
>
>
>
> joe
>


--
WenSui Liu
A lousy statistician who happens to know a little programming
(http://spaces.msn.com/statcompute/blog)