Date:         Wed, 11 Dec 1996 16:44:35 -0800
Reply-To:     Donald Peter Cram <doncram@LELAND.STANFORD.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Donald Peter Cram <doncram@LELAND.STANFORD.EDU>
Organization: Stanford University, CA 94305, USA
Subject:      Re: Bootstrap SEs in SAS -- help needed

In article <58hd3i$og7@sjx-ixn9.ix.netcom.com>,
Michelle Malkin <malkin1@ix.netcom.com> wrote:
>Can someone refer me to resources on bootstrapping in SAS. I need to
>get bootstrapped estimates of Standard Errors from a complicated
>regression.
>
 
Paul Thompson provided compact, efficient SAS code for bootstrapping
in his SUGI 21 (1996) paper "A Tutorial on Bootstrapping in the SAS
System".  I recall he also posted the code in this forum (check the
archive using commands in the SAS-L FAQ), although I still recommend
getting and reading his article.  The SUGI 21 proceedings may be
available at your library;  you can buy them from SAS Institute for $50
or so yourself.
 
What you want to avoid is creating a separate resample dataset for
each of N bootstrap iterations, and looping through data steps and
PROCs.  Instead you want one large dataset consisting of N resamples,
which you analyse once, say in PROC REG using a BY variable that is
the resample number.  Also you want to take care to save your seed
values for the random selection, and to take care of more
complications which Paul did well.  Worth >>> $50 of your time.
 
hope this helps
Don Cram
 
 
 
--
doncram@gsb-ecu.stanford.edu
http://www-leland.stanford.edu/~doncram