Date: Thu, 23 Oct 2008 17:17:26 -0700
Reply-To: tanwan <tanwanzang@YAHOO.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: tanwan <tanwanzang@YAHOO.COM>
Subject: Re: calculating means in a subsampled survey design
Content-Type: text/plain; charset=ISO-8859-1
Yes, your code works correctly.
On which variable to put in the cluster statement: you just need to
put the one that is 'higher' in rank between Treatment and Area.
Imagine you work for your home state Dept. of Education, and you are
studying something about school kids. You might start with a list of
all school districts and group them into either RURAL or URBAN and
take a sample from each. This rural/urban variable would be your
STRATA variable on proc surveymeans or any other surveyproc. Imagine
further that from the sampled school districts you take a sample of
schools, and from the sampled schools you take a sample of classes/
grades, and from the selected classes you take a sample of pupils.
Notice here that pupils are nested within classes, classse are nested
into schools, and schools are nested in school districts.
So which of these variables goes on the CLUSTER statement of the
Because school district is on top of the perch, it does.
Since you know your study better than I, you will then choose between
Treatment and Area to plug into the CLUSTER statement.