Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Re: how to select model in xtpcse , corr(ar1)


From   Christopher Baum <baum@bc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: how to select model in xtpcse , corr(ar1)
Date   Thu, 4 Dec 2008 14:53:04 -0500

< >
Ghislain said
I want to compare two models one made with : xtpcse x y z,corr(ar1) the other with : xtpcse x y w,corr(ar1) xtpcse doesn't return AIC nor BIC nor e(ll),

Martin suggests looking at r^2. AIC and BIC are criteria that will penalize for non-parsimonious models. As both of the models you specify have exactly the same length, they would not be useful. They are not estimated with maximum likelihood, so there is no e(ll). [OLS is not estimated with MLE either, but -regress- does provide e(ll)]. As r^2 is perfectly comparable for models of the same length, r^2 would seem to make sense as a comparison.

I would worry more about the notion that if each of these models does a good job, with a significant z or w respectively, then there is surely the risk that they are both misspecified versions of a model that contains both. Can you indeed rule out w appearing in the first model, or z in the second?


Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index