Date:         Tue, 13 Aug 2002 09:19:41 -0700
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: %DO %WHILE confusion
Comments: To: greenberg@UCDAVIS.EDU
Content-type: text/plain; charset=us-ascii

Jonathan Greenberg <greenberg@UCDAVIS.EDU> wrote [in small part]:
> Part IV of the great SAS experiment (incidentally, sorry for all the
posts
> -- I started learning SAS yesterday, so I am a total newbie)...

Well, in that case, I am quite impressed at the speed you're
picking all of this up.  Keep up the good work!

I'm eliding all of your code, since others have given you some
excellent advice already.  I, however, am more concerned with your
statistical approach.  I'm really concerned that, at the end of your
processing, you will *NOT* have a statistically valid test of the
power you think.  Iteratively punting leverage points and outliers
has a number of risks, which I and others discussed yesterday.

Are you absolutely sure that this is the way you have to go with these
data?

Are you sure that you really have one single population from which all
your data points flow?

You are fitting a very simple model.  Are you sure that your model
is really appropriate, given the amount of stat-fiddling you seem to
need?  Are you sure that linear regression is really appropriate?

I realize that I have thrown a lot of questions out, but I see a lot
of room for concern.  In my years of consulting, I have found that
it is always easier to fix the conceptual model at the beginning
than to have to re-work things after a paper has been submitted or
presented, and outside interests have jumped all over one's, errm,
case.  :-)

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