Date:         Mon, 12 Aug 2002 13:39:52 -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: Newbie MERGE question
Content-type: text/plain; charset=us-ascii

Jonathan Greenberg <greenberg@UCDAVIS.EDU> replied [in part]:
> A semi-minor point -- this is going to be ultimately applied to a
dataset
> with 700,000 + observations, and will have the analyze --> delete
> observations with too high a DFBETA--> check current stats--> repeat,
so I'm
> worried the MERGE step will take up too much memory/processing time --
there
> is no way to extract the observation # from the dfbeta set which
violate the
> rule (e.g. Over a certain threshold), and then use that to delete the
> original data, instead of merging each time?

So you want to look in an iterative fashion at points of influence in
your model.  Are you sure this is a good idea?  If someone is ordering
you to program this, then you're stuck.  But if this is up to you, then
you may want to look at robust or resistant methodologies instead.

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