Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: RE: st: significance of model improvement using log pseudolikelihood?


From   Maarten buis <[email protected]>
To   Bart Vanneste <[email protected]>
Subject   Re: RE: st: significance of model improvement using log pseudolikelihood?
Date   Fri, 30 Dec 2005 00:23:34 +0000 (GMT)

dear Bart,

Testing for improvement in model between nested models fit is the same as testing constraints. In
your case a very easy one, but sometimes the constraint can be very complicated as Stas pointed
out. In many cases, especialy with easy constraints, looking at the test as a test of constraints
is more substantively interesting than looking at the test as a test of improvement in fit. Our
substantive interest usualy lies with wether variables x, y, and z have an effect or not (and if
so how large the effect is), and not in whether model 1 fits the data beter than model 2. So my
advise is to always keep the question you want to answer in mind, and choose your test and
interpretation of the test accordingly. In many cases you will find the "test of constraints"
interpretation more meaningfull than the "improvement of fit" interpretation, regardless of
whether the difference between models is one or more variables / number of constraints is one or
more. 

A more specific answer to your question is that the p-value you reported is already from the wald
test. So it would indeed be double to report the same test twice.

HTH,
Maarten

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting adress:
Buitenveldertselaan 3 (Metropolitan), room Z214

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


--- Bart Vanneste <[email protected]> wrote:

> In my case model 2 contains just one variable more than model 
> 1.  Testing the increase in model fit would then be equivalent 
> to testing whether the coefficient is different from zero. 
> 
> From a practical viewpoint, is there any value in separately 
> reporting  the significance in model fit given that I already
> report the P-values of the individual coefficients? Or does it 
> only make sense when at least 2 variables are added in 
> subsequent models?



		
___________________________________________________________ 
Yahoo! Exclusive Xmas Game, help Santa with his celebrity party - http://santas-christmas-party.yahoo.net/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index