Date:         Mon, 29 Aug 2005 14:44:30 -0700
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: SAS Proc Logistic - Stepwise : how to fix a variable to be
              included in all models
In-Reply-To:  <MC4-F17zRdKotyAfWlJ00136821@mc4-f17.hotmail.com>
Content-Type: text/plain; format=flowed

paul@WUBIOS.WUSTL.EDU sagely replied:
>one of the advantages of the SAS system is that it includes the macro
>system.  Using macros, it is simple to write your own stepwise
>regression approach, in which variables are forced in, and then other
>variables are allowed to stay in or drop out.  Many statisticians,
>myself included, do things like this and don't find them objectionable -
>   others find these problematic as well.

This is a very good point.  A *key* issue here is the use of appropriate
statistical methodology in the process.  A statistician like Paul is not
going
to throw 5,000 varaibles into the hopper, close his eyes, press the
STEPWISE button, and automagically trust the results.

He's going to evaluate variables first.  He's going to look for problems
that
lead stepwise (and other seleciton methodologies) astray, and address them
first.  He's going to perform regression diagnostics.  He's going to look at
the results and evaluate them in terms of "does this make any sense based
on current theory".  All these things need to be done.  I know Paul will do
these things, and use this tool carefully.  I just don't feel comfortable
with
everyone else on the planet using stepwise.  :-)

Stepwise selection is a tool which can be misused.  But the same thing has
been said about every SAS proc at one time or another.  Back in the 1980's
I occasionally worked with a Crotchety Old Statistician (tm) who said more
than once that SAS was too darn user friendly, making it too easy to misuse.
He wanted SAS to start up with a pop quiz in the splash screen, and if you
couldn't answer the stat questions well enough, the program should log you
off.  :-)  :-)

Still working on my 'Crotchety Old Statistician' certification,
David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/