bartlett correction to the likelihood ratio test in multilevel analysis

[Solange Sardan]

Good morming everyone!
I'm working in a multilevel model with two level. I have the problem that the number of groups is too small and the wald test is not the indicate to test the significance of the variables. For this reason a read i should apply a bartlett correction to the likelihood ratio test, but i'm not sure how should i do it. Is there an specific command to fo it?
Thank you for your time. Greetings

[Stephen Jenkins]

Please provide full literature references regarding the "Bartlett correction" (see the Forum FAQ about this!). What is currently available in built-in Stata can be browsed by looking at help mixed and following relevant links to the Manuals

[Solange Sardan]

Sorry, i think i didn't explain myself. I am testing the significance of the fixed part of my multilevel model, and i have the special case where the number of groups is too small, therefore i read that in that case one should use a bartlett-correction to the likelihood ratio test, so type I error rates improve.

[Stephen Jenkins]

Solange, you have simply repeated yourself. I asked for a literature reference regarding Bartlett corrections. Also, please do read the Forum FAQ about providing full references, and how to post questions effectively. In addition, you could google on "Stata Bartlett correction" and follow links. Finally, if the number of groups is "too small", I don't see how some post-estimation 'magic' can provide a more reliable answer. [It's this sort of scepticism that needs to be met with evidence from published literature!]