Date:         Fri, 20 Jun 2008 11:55:58 -0400
Reply-To:     Sigurd Hermansen <HERMANS1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Sigurd Hermansen <HERMANS1@WESTAT.COM>
Subject:      Re: Bootstrap sampling question
Comments: To: joylelefei@yahoo.com
In-Reply-To:  <689189.83130.qm@web45107.mail.sp1.yahoo.com>
Content-Type: text/plain; charset="iso-8859-1"

Joy:
Are you bootstrapping posts to SAS-L? That might explain the eight replicates of your question.
S

-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of Joy Fei
Sent: Friday, June 20, 2008 11:15 AM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Bootstrap sampling question


Hi All,

I want to generate bootstrap samples based on the following original sampling:

                                          Testing Results at Instrument 1            Testing Results at Instrument 2 Sample ID            SITE1            SITE2            SITE3   SITE4                           SITE1

1                            +            -            +            +                                  + 2                      -            +            +            +                                  + 3                      -            +            +            +                                  + ..

50                     +            +            +            +                                  +

The ideal is that I want to compare Instrument 1 to the Instrument 2 but there are 4 results for Instrument 1. Sample ID is the patient blood samples and they are the same sample for all sites for each ID. Because the same 50 samples are tested at all sites, and I need to combine results for all sites for Instrument 1 and perform a comparison study with the instrument 2, a bootstrap re-sampling might work well and enough. My original thought is to re-sample 200 (50 * 4 sites) IDs (but only 50 different samples including all 4-site testing results at Instrument 1) and generate a bootstrap sample with 50 IDs each time then do a comparison to the Instrument 2. Can I use proc surveyselect and what is the right way? I appreciate your any suggestions.

Thanks a lot,
Joy