Date:         Mon, 27 Oct 2008 10:09:09 -0700
Reply-To:     "sophe88@yahoo.com" <sophe88@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "sophe88@yahoo.com" <sophe88@YAHOO.COM>
Organization: http://groups.google.com
Subject:      Sampling or not, this is the question
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

Hi,

Hope to hear some input on this question. This is more like a thought
question. At frequency table

                                                       variable2
Variable 1      x                       y
z                              Row total
0               366,700,892     256,364,259     592,514,321           1,215,579,472
1                       0              265,412,326       69,512,786             334,925,112
2                       0              26,598,741               263,578,912             290,177,653
3                       0             61,037,890                689,478,021              750,515,911
 total        366,700,892            609,413,216               1,615,084,040
2,591,198,148

This table is from merging 2 data sets by simple "if A & B ", by ID.
Data set A has variable 1 and B has variable 2.

This table used  2006 data. Now we need to update it with 2008.
Problem is 2008 has much larger counts

So one suggested: we should do a random sample, say ranuni ( 92929)
<0.25 on both a and b.

I frowned : if so, each ID has 25% chance to be picked. One ID could
be excluded because it belongs to the 'unlucky' 75%, not because it is
not matched to the other table.

So, to sample or not to sample? Thanks.

PD