Date: Sat, 25 Jun 2005 12:35:57 +0200
Reply-To: Pierre Gaudebout <firstname.lastname@example.org>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Pierre Gaudebout <pierre.gaudebout@GMAIL.COM>
Subject: Re: comparison with weighted percentages (complement)
Content-Type: text/plain; charset=ISO-8859-1
Some elements concerning the sampling design.
The last 150 patients discharged from hospital as from one D-day
(transverse survey) are questioned on their stay. Twenty five
hospitals are concerned with the
investigation but they do not form a homogeneous group in term of
capacity (there is approximately a ratio from 1 to 25 between smallest
and the largest hospital as regards the number of annual admissions).
It would be too expensive to carry out this investigation with
stratified samples with proportional allocation. We have to choose
samples with equal allocation but with "rectification" by giving a
different weight to hospitals (this
weight is calculated like the ratio between the number of
annual admissions of the hospital and that of the whole of the studied
Concerning the choice to modify the coefficients for the test, it is
simply a question of redefining the weight of each hospital of the
group (H2, H3, H4 and H5, to take again the example).
Personally I am not convinced perfectly of the utility to balance but
in fact the choice was endorsed by statisticians more senior than me.
In all the cases thank you to David Cassell for your valuable comments