Date:         Fri, 5 Dec 2008 15:52:45 -0600
Reply-To:     Mary <mlhoward@avalon.net>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Mary <mlhoward@AVALON.NET>
Subject:      Re: PROC LOGISTIC: how to run with data with frequencies?
Comments: To: "./ ADD NAME=Data _null_," <iebupdte@GMAIL.COM>
Content-Type: text/plain; charset="iso-8859-1"

Thanks so much.

-Mary
  ----- Original Message ----- 
  From: ./ ADD NAME=Data _null_, 
  To: SAS-L@LISTSERV.UGA.EDU 
  Sent: Friday, December 05, 2008 3:44 PM
  Subject: Re: PROC LOGISTIC: how to run with data with frequencies?


  767  proc logistic;
  768     class allele;
  769     model disease/freq = allele;
  770     run;


  On 12/5/08, Mary <mlhoward@avalon.net> wrote:
  > I know I've seen it before, but can't seem to find this.
  >
  > Is there a way to run PROC LOGISTIC using frequencies instead of
  > observation-level data?   I've got data like this:
  >
  > data test;
  > informat allele $1.;
  > infile cards;
  > input allele disease freq;
  > cards;
  > A 0 353
  > A 1 280
  > B 0 424
  > B 1 323
  > run;
  >
  > That I want to run a model of disease=allele with.
  >
  > Does anyone know how to do this?
  >
  > -Mary
  >
  > ----- Original Message -----
  > From: Cornel Lencar
  > To: SAS-L@LISTSERV.UGA.EDU
  > Sent: Friday, December 05, 2008 3:17 PM
  > Subject: Re: tphreg with time-varying covariates
  >
  >
  > Hi Deb,
  >
  > Try this approach in your model statement:
  >
  > do i = 1 to dim(mxs);
  >       if mxs{i} ne . and mxdts{i} ne . and mxdts{i} lt evdt then t_cd =
  > mxs{i};
  >       t_cd4cat=1*(tcd ge 0 and t_cd lt 200) + 2*(t_cd ge 200 and t_cd le
  > 349) + 3*(t_cd ge 350 and t_cd le 499) + 4 *(t_cd ge 500 and t_cd le 649) +
  >       5*( t_cd ge 650);
  > end;
  >
  >
  > In your initial code, the variable t_cd4cat was built outside the array so
  > the model actually didn't see it. Now I think it should work.
  >
  > Cheers,
  >
  > Cornel
  > -----Original Message-----
  > From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On
  > Behalf Of Kevin
  > Viel
  > Sent: Friday, December 05, 2008 1:00 PM
  > To: SAS-L@LISTSERV.UGA.EDU
  > Subject: Re: tphreg with time-varying covariates
  >
  > On Thu, 4 Dec 2008 09:29:24 -0600, Deborah Wentworth <debby@CCBR.UMN.EDU>
  > wrote:
  >
  > > Greetings -
  > >
  > > I am trying to run my first tphreg model.  It's complicated by the fact
  > > that I have a time-varying covariate.
  > >
  > > I cannot find an example of the syntex.  The log tells me that "variable
  > > T_CD4CAT not found".
  > >
  > > My code is below.  Can someone tell me if I have the incorrect syntax for
  > > what I am trying to do?  Because if's my first stab at tphreg ever, I
  > > may not be correctly interpreting use of the contrast statement.
  > >
  > > Your help would be much appreciated!
  > >
  > > Thanks, Deb
  > >
  > > proc tphreg data=events;
  > > class gpa t_cd4cat / ref=first ;
  > > model t2dth * dth(0) =  gpA t_cd4cat gpA*t_cd4cat / risklimits;
  > >
  > > contrast 'GpA CD4 cat 1' gpa 1 / estimate=exp;
  > > contrast 'GpA CD4 cat 2' gpa 1 gpa*t_cd4cat 1 0 0 0 / estimate=exp;
  > >
  > > array mxdts{*}  mxdt1-mxdt32;
  > > array mxs{*}  cd1-cd32;
  > > t_cd = cd4bl;                     /* initially set to baseline */
  > >
  > > do i = 1 to dim(mxs);
  > >  if mxs{i} ne . and mxdts{i} ne . and mxdts{i} lt evdt then t_cd = mxs
  > >
  > {i};
  > > end;
  > >
  > > if 0 le t_cd lt 200 then t_cd4cat=1;
  > >  else if 200 le t_cd le 349 then t_cd4cat=2;
  > >  else if 350 le t_cd le 499 then t_cd4cat=3;
  > >  else if 500 le t_cd le 649 then t_cd4cat=4;
  > >  else if t_cd ge 650 then t_cd4cat=5;
  > >
  > > run;
  > >
  >
  >
  > Hi Deborah,
  >
  >  Interestingly, we have similar questions:
  >
  > PHREG with time-varying covariates
  > http://www.listserv.uga.edu/cgi-bin/wa?A2=ind0812a&L=sas-l&F=&S=&P=44738
  >
  >  I am curious about this line:
  >
  > > if mxs{i} ne . and mxdts{i} ne . and mxdts{i} lt evdt then t_cd = mxs{i};
  > >
  >
  > In your MODEL statement:
  >
  > > model t2dth * dth(0) =  gpA t_cd4cat gpA*t_cd4cat / risklimits;
  > >
  >
  > the time seems to be t2dth.  As I understand PHREG, and for your safety
  > assume that I do not understand it correctly, the time of interest is
  > coming from the MODEL statement.
  >
  >  Also, you *might* be able to omit the baseline assignment statement if
  > you reconstruct your array to include it as the first value with the
  > corresponding time of 0?
  >
  > HTH,
  >
  > Kevin
  >