Date:         Wed, 21 Apr 1999 14:22:09 -0400
Reply-To:     "Rul-lan, Katherine" <Katherine.Rul-lan@BASES.ACNIELSEN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         "Rul-lan, Katherine" <Katherine.Rul-lan@BASES.ACNIELSEN.COM>
Subject:      Significance Testing for Partial Overlapping Groups
Content-Type: text/plain; charset="iso-8859-1"

Hello All.
 
I am sometimes asked to perform a statistical significance test on partially
overlapping groups.  For example, do the buyers of Ziploc differ
significantly from the buyers of Hefty One-Zip on some variable (could be
satisfaction or number of units bought)?  The issue is that I have partial
overlapping samples.  I have consumers that are in both groups (hence
dependent) and those that are only Ziploc buyers or only Hefty buyers (hence
independent).  (Note: one group is not a subset of the other.)  Has anyone
else ever had to deal with this?  I want to take advantage of the fact that
part of the consumers are the same and others are not in calculating the
sample variance.
 
I do have some equations written to do this but have never had the chance to
program it.  Does anyone have anything in SAS (I don't have IML) or Excel
that they wouldn't mind sharing?
 
Thanks
Katherine Rul-lan
ACNielsen BASES
Diagnostic Consulting
606-655-6220