What's wrong with the use of -qic-?
I guess na�ve likelihood ratio is minus two times the loglikelihood difference. You can have it from -bicdrop1- from ssc. Plase note the warning "not yet tested for xtgee (may not be correct)". Ask the author. Wald may be based on the Wald test that is reported above the table of the results of -xtgee-, on the right. -predict, score- calculates the equation-level score, but I have no idea about the score test.
Nicola
P.S. More persons may engage the discussion if you provide full reference.
At 02.33 08/11/2007 -0500, "Ed Levitas" wrote:
>Statalisters,
>
>I am trying to perform estimations using XTGEE ( family(normal), link(identity), corr (ar1)) and want to compare restricted vs. unrestricted models. Hardin & Hilbe (2003; 169-174) discuss Wald, na�ve likelihood ratio, and score tests. From their discussion, I believe I can use these tests to compare, for example, a main effects model vs. a main effects plus interaction model. However, I cannot figure out how to construct these three types of tests using STATA. Can someone offer suggestions?
>
>Thanks in advance
>Ed
*
* 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/
