Date:         Mon, 17 Dec 2007 15:45:15 -0600
Reply-To:     Mary <mlhoward@avalon.net>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Mary <mlhoward@AVALON.NET>
Subject:      Re: Performing thousands of tests automatically
Comments: To: "data _null_," <datanull@GMAIL.COM>
Content-Type: text/plain; charset="iso-8859-1"

What about with a single independent (X) variable is changing?  In most of these models the dependent variable is static (some indicator of whether the person has the disease or not), and each run would need to have a different X variable, in the case of genetics, the individual SNP that would change, plus perhaps other variables in the model like age. 

-Mary
  ----- Original Message ----- 
  From: data _null_, 
  To: SAS-L@LISTSERV.UGA.EDU 
  Sent: Monday, December 17, 2007 3:37 PM
  Subject: Re: Performing thousands of tests automatically


  I forgot about that.  After the correct transpose there should be ONE
  and ONLY ONE column var created COL1 it becomes the Y var in the
  model.

  by _NAME_;
  model col1 = independents.

  I would probably rename it Y.

  On Dec 17, 2007 3:26 PM, Mary <mlhoward@avalon.net> wrote:
  >
  > I don't think this will work since the mixed model statement would have to
  > have the name of the variable name in it, and thus a by _name_ statement
  > wouldn't fix the model statement- that's why I suggested loading up a data
  > set with the variables and putting each variable into a macro variable each
  > time.
  >
  > proc mixed;
  >    model dependent_var=independent_var;
  >
  > changes to:
  >
  > proc mixed;
  >    model dependent_var= &independent_var;
  >
  >
  > -Mary
  >
  > ----- Original Message -----
  > From: data _null_,
  > To: SAS-L@LISTSERV.UGA.EDU
  >
  > Sent: Monday, December 17, 2007 2:23 PM
  > Subject: Re: Performing thousands of tests automatically
  >
  > Something like this untested.
  >
  > proc transpose data=wide out=tall;
  >    by <unique> <right hand side variables>; * such that each
  > observation produce VARs observations.;
  >    var var:; *or some other list that works for you.;
  >    run;
  > proc sort data=tall;
  >    by _name_;
  >    run;
  > proc mixed...
  >    by _name_;
  >    model
  >    <other mixed statements.>
  >    run;
  >
  > Test on a small batch of variables, to debug.
  >
  >
  >
  > On Dec 17, 2007 2:15 PM, Ron Do <rondo@hotmail.com> wrote:
  > > Hi,
  > >
  > > I would like to perform regression analysis using proc mixed for over a
  > > hundred thousand different independent variables (ie. var1 to var100000).
  > > Can I do this without having to change the variable and repeatedly click
  > the
  > > "run program" button every time?
  > >
  > > Thanks for your help,
  > > Ron
  > >