Date:         Tue, 4 Apr 2006 20:32:47 -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: Vuong's test
In-Reply-To:  <200604050323.k34M2Zrm014810@mailgw.cc.uga.edu>
Content-Type: text/plain; format=flowed

art297@NETSCAPE.NET replied:
>Take a look at: http://xrl.us/kphd

Yes.  Dale's posts are always instructive.

However, as I pointed out in one of the 'please give me code to run
Vuong's test' threads on SAS-L, naive usage of Vuong's test has
potential drawbacks.  I said:

==========================================
Let me also point out that Vuong's statistic has to be fiddled with if the
model degrees of freedom are different between the models, and people
have questioned the validity of the results for small or moderate sample
sizes.

Vuong's test (and some others, like Clarke's 'non-parametric' test) is based
on the comparison of the Kullback-Liebler Information Criterion for more
than one model.  If your models are not appropriate for this, then you
may want to refrain from this.

And finally, the over-comparison of a bunch of different models is a
traditional way to get yourself into a really miserable bind, since
eventually you may end up picking a model which does a great job of
fitting the noise in your data, and hence will NOT provide predictive
capabilities in future, and will NOT provide realistic parameter estimates.
Don't get carried away.
==========================================

That advice still holds.

So a relevant question is: what does Ivy want to do with Vuong's test,
and what is her overall goal?

HTH,
David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
Is your PC infected? Get a FREE online computer virus scan from McAfeeŽ
Security. http://clinic.mcafee.com/clinic/ibuy/campaign.asp?cid=3963