=========================================================================
Date:         Mon, 24 Jul 2006 05:52:15 -0700
Reply-To:     Albert-jan Roskam <fomcl@yahoo.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Albert-jan Roskam <fomcl@yahoo.com>
Subject:      weighted numbers
In-Reply-To:  <945D579F76FD5741A043511E6FE99B570A43FE74@TLRUSMNEAGMBX32.ERF.THOMSON.COM>
Content-Type: text/plain; charset=iso-8859-1

Dear all,

I want to calculate the relative proportions of people
within educational categories, stratified by sex and
country. I use a standardized weight that corrects for
sample selection bias.  Could anybody tell me
why-oh-why the two methods listed below yield
(slightly) different results? The N function of the
aggregate counts the weighed number of cases within
the break group. So why are the AGGREGATE results
identical to the output I get using SUMMARIZE with an
UNweighted sample? Am I missing something?

Thanks a lot in advance!

Albert-Jan


weight by w_plus.

** Method 1.
compute dummy = 1.
summarize tables = dummy by sex by country3 by educat
/   cells = npct (educat).

** Method 2.
aggregate outfile = "d:\temp\tmp.sav" / break = sex
country3 educat / n = n.
aggregate outfile = * / break = sex country3 / n_tot =
n.
match files table = * / file = "d:\temp\tmp.sav" / by
= sex country3.
compute pct = (n / n_tot ) * 100.
exe.

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around