Date:         Thu, 9 Nov 2000 08:42:24 -0500
Reply-To:     John Jones <jonesj@PHARMARESEARCH.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         John Jones <jonesj@PHARMARESEARCH.COM>
Subject:      Re: PHREG and dummy variables
Comments: To: Luca Boni <boni@CBA.UNIGE.IT>
In-Reply-To:  <3.0.6.32.20001109122717.00909480@ermes.cba.unige.it>
Content-Type: text/plain; charset="us-ascii"; format=flowed

Hi Luca,

There does not appear to be a CLASS statement for PHREG.  You'll need to
create indicator variables in a data step.  Pick a reference treatment and
omit its respective variable from the PHREG model.

Data Shmata;
    set MyTreatment;
    TrT1=0;
    TrT2=0;
    TrT3=0;
           if (TrT='PLACEBO') then TrT1=1;
    else if (TrT='LOW')         then TrT2=1;
    else if (TrT='HIGH')         then TrT3=1;
run;

Proc PHReg data=Shmata;
  model response = TrT2 TrT3;
run;

At 12:27 PM 11/9/00 +0100, Luca Boni wrote:
>Hi all:
>
>Question about SAS V8...
>
>Is there any statement (such as CLASS in the PROC GLM) to define a
>classification variable inside the PROC PHREG ? (i.e. the treatment
>variable has three levels and I would like to apply a stepwise selection
>method)
>
>Thanks
>
>Luca
>
>
>
>
>        \\\|///
>      \\  - -  //
>       (  @ @  )
>----o00o-(_)-o00o----------------------------------------------------
>
>Luca Boni, MD
>Unit of Clinical Epidemiology and Trials
>Advanced Biotechnology Center-National Institute for Cancer Research
>GENOVA-ITALY
>Tel.: +39 010 5737477
>Cellular Phone: +39 335 6777997
>E-mail: boni@cba.unige.it