Date:         Tue, 16 Nov 2004 16:37:41 -0500
Reply-To:     Bob_Abelson@HGSI.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Bob Abelson <Bob_Abelson@HGSI.COM>
Subject:      Re: Random selection
Content-Type: text/plain; charset="iso-8859-1"

The easiest way, although far from the most efficient, would be to assign 
a random number to each observation, sort by that number, and take the 
first 150 of them.

data sample;
        set yourdata;
        retain seed 6345789;
        randno = uniform(seed);
run;

proc sort data=sample;
        by randno;
run;

data sample;
        set sample;
        if _n_ <= 150;
        drop randno;
run;

I only recommend this with relatively small datasets, like yours. Larger 
datasets need a more efficient technique.

Hope this helps.

Bob Abelson
HGSI
240 314 4400 x1374
bob_abelson@hgsi.com

"Coffee falls into the stomach ... ideas begin to move, things remembered 
arrive at full gallop ... the shafts of wit start up like sharp-shooters, 
similies arise, the paper is covered with ink ..." 
-- Honore de Balzac 




Myra <myrabarnes@YAHOO.COM>
Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
11/16/2004 04:16 PM
Please respond to Myra

 
        To:     SAS-L@LISTSERV.UGA.EDU
        cc: 
        Subject:        Random selection


I have a dataset of about 2800 observations & I need to randomly select 
150 of them & output them into a new dataset. How would I go about doing 
this?

Thanks for any help.

~Myra~



---------------------------------
Do you Yahoo!?
 The all-new My Yahoo! ? Get yours free!