Date:         Thu, 7 Nov 2002 09:46:30 -0800
Reply-To:     Cassell.David@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <Cassell.David@EPAMAIL.EPA.GOV>
Subject:      Re: "None of the Above" Analysis - Addendum (LONG)
Comments: To: Peter Flom <flom@NDRI.ORG>
Content-type: text/plain; charset=us-ascii

Peter Flom <flom@NDRI.ORG> wrote a long reply, chock full
of wisdom:

I just want to agree with every one of Peter's points,
although mentioning in passing that the meter doesn't seem
quite right in one or two lines of his poem.  :-)

The original poster has gotten himself into a bind with the
design of the study, and doesn't know how to allocate the
"none of the above" ratings across the full N categories.

My feeling is that he cannot.  There is no way to do so
which *fully* takes into account the problem of uneven
non-response, i.e., different combinations of categories
would get different levels of "none of the above" entries,
in such a way that these cannot be relaiably filled in
when moving to larger sets of categories.  It seems to me
that Will is stuck having to perform some analyses on the
full set of categories *without* including the "none"
category.  There isn't any way he can guess how to distribute
the "none" category properly across all N categories, and
making guesses will just introduce errors that may swamp or
massively distort the results.

David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician