Date:         Wed, 13 Jun 2001 09:10:58 -0400
Reply-To:     Philip Whittall <philip.whittall@UNILEVER.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Philip Whittall <philip.whittall@UNILEVER.COM>
Subject:      Re: Indicator Variables in Regression

Dear Henry,

             Basically all the literature about
analysing non-orthogonal designs holds. Take a
good look about stats for colinearity, the 4
types of sums of squares and the correlation
of the parameter estimates. What is interesting
about your data is not that the categories overlap,
but that they form a hierarchy. That is if you have
a doctorate you almost certainly have some of the
other qualifications too, same for masters. It all
depends on the purpose of your analysis, but you might
want to consider recording the highest qualification held
for some purposes,

Philip