Date:         Wed, 8 Aug 2007 11:06:33 -0700
Reply-To:     sounpra@YAHOO.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Song <sounpra@YAHOO.COM>
Organization: http://groups.google.com
Subject:      Re: bootstrap simulation
Comments: To: sas-l@uga.edu
In-Reply-To:  <7367b4e20708071409l5927ebedm56698ff127627e52@mail.gmail.com>
Content-Type: text/plain; charset="us-ascii"

On Aug 7, 4:09 pm, datan...@GMAIL.COM ("data _null_;") wrote:
> You want to use SURVERSELECT for this and also read Cassells.  If you
> really think you need the original data stuck to the front of  the
> boot strap sample you can create a view as below.
>
> You may want to use a "seed" so you can reproduce the result if need be.
>
> data example;
>    do id = 1 to 10;
>       group1 = rannor(1234);
>       output;
>       end;
>    run;
> proc surveyselect Seed=177161001 data=example method=urs reps=10
> out=work.boot outhits rate=1;
>    run;
> data work.bootV / view=work.bootV;
>    if 0 then set work.boot(keep=Replicate);
>    Replicate = 0;
>    do until(eof);
>       set work.example end=eof;
>       output;
>       end;
>    eof = 0;
>    do until(eof);
>       set work.boot end=eof;
>       output;
>       end;
>    stop;
>    run;
> proc print;
>    run;
>
> On 8/7/07, Song <soun...@yahoo.com> wrote:
>
>
>
> > Hi All,
>
> > I have the following code:
>
> > /********************************************/
> > data example;
> >    do i = 1 to 10;
> >        group1 = rannor(0);
> >        output;
> >    end;
> >  keep group1;
> > run;
>
> > data boot;
> >  do sample = 1 to 10;
> >     do i = 1 to nobs;
> >            x = round(ranuni(0) * nobs);
> >                 set example
> >          nobs = nobs
> >          point = x;
> >        output;
> >       end;
> >    end;
> >  stop;
> > run;
> > /*****************************************/
>
> > What I would like to do is simulate my rannor data multiple times and
> > perform abootstrapto each of those simulated dataset.  Can anyone
> > suggest how I might be able to combine the two data step into one with
> > the additionalsimulationvariable?
>
> > Thank,
>
> > Song- Hide quoted text -
>
> - Show quoted text -

Thank you for everyone's response.  The paper by Cassell et. al was
helpful in using PROC SURVEYSELECT.

However if I wanted to simulate 1000 sets of 20 observations from a
normal distribution:

data mydata;
seed1 = 123
 do sim = 1 to 1000;
    do i = 1 to 20;
        group1 = rannor(seed1+ sim);
        output;
    end;
  end;
  drop seed1;
run;

Then I would like to perform 1000 bootstrap reps on each of the
individual simulated dataset of 20 observations.  How can I modify the
following code to account for my data?

proc  surveyselect data = mydata    out = boot
     seed = 321
     method = urs  sample = 1   outhits
     rep = 1000;
run;

Thanks,

Song