Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: lrtest: negative values


From   rgutierrez@stata.com (Roberto G. Gutierrez, StataCorp)
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: lrtest: negative values
Date   Fri, 04 May 2007 08:29:02 -0500

Christoph Wunder <christoph.wunder@sowi.uni-bamberg.de> asks:

> I am testing for the significance of a random intercept in a two-level model
> using lrtest (after xtmixed). Although I use exactly the same dataset and
> the same set of explanatory variables (i.e., the models differ only in the
> level-2 random intercept), the log-likelihood of the unrestricted model is
> smaller than of the restricted one. lrtest calculates a negative value of
> the test statistic.

> How can one interpret the result? Any suggestions?

[output deleted]

Your models are not nested (in the model sense), and thus you should not 
expect the ordinal relationship in log-likelhoods associated with nested
models.

Your "restricted" model has random effects at the levels of variable -level1-.

Your "unrestricted" model has random effects at the levels of variable 
-level2-, and random effects at the levels of variable -level1- nested within
-level2-.

As such, your restricted model cannot be considered a special case of the
unrestricted model.  Perhaps you meant instead to have -level2- nested within
-level1- for the second model, in which case you could then apply an LR test.
If this is so, simply switch the ordering of the level specifications in your
call to -xtmixed-.  Recall, the ordering of nesting goes left to right from
biggest groupings to smallest groupings.

--Bobby
rgutierrez@stata.com
*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index