st: MI and Mixed Effects Models

[Kris Anderson <andersok@reed.edu>]

I have run a number of models using mi estimate: xtmelogit. I am having 
trouble getting some information when using the mi option.

1. When I run xtmelogit without mi, I get significance values for the 
random effects parameter (chi bar squared). I don't get this when using 
MI. How do I estimate whether the random effects parameter is 
significant using MI? On a related note, how can I estimate the 
interclass correlations?

2. I have significant 2 way interactions in these models. Is there a way 
to probe these in Stata taking the random effects into account? I have 
searched the documentation without luck.

I have found a website that allows for probing these interactions using 
R. However, I am having difficulty pulling together all the coefficients 
I need using Stata. The program requires that I have get the asymptotic 
covariance matrix of the regression parameters. How can I get these from 
Stata? estat only has options for recovariance (it doesn't seem to work 
when I use MI anyway).

I would appreciate any help or suggestions. I have been combing the 
documentation and the web without luck.

Thanks,
Kris
*
*   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/