st: permute with multi-level models

["Jessica Gottlieb" <jgott@stanford.edu>]

Hi, 

I am trying to run a Monte Carlo simulation on a mixed model regression to
infer standard errors and p values. 

I am using the permute command after running a multi-level model with
xtmixed (see below).  The treatment variable t has 3 values {0,1,2} and I am
interested in estimating the coefficients using i.t.  The data is
multi-level in which individuals are nested within villages which are nested
within communes (reflected in the random effects).  

program pxtmixed, rclass
	xtmixed trust i.t || commune: || village:
	matrix x=e(b)
	return scalar t1 = x[1, 2]
end
permute t r(t1), reps(1000): pxtmixed

I have 2 major questions: 
1. I want to permute t across communes such that the value of t is uniform
for all individuals within a given commune.  Does anyone have thoughts on
how to do this?
2. I am interested in estimating p values for 2 coefficients, but when I try
to return a matrix rather than a scalar (see below), the command doesn't go
through.  I get an error that says "weights not allowed".

program pxtmixed, rclass
	xtmixed trustcouncil i.t || commune: || village:
	matrix x=e(b)'
	return matrix coeff = x[2..3, 1]
end
permute t r(coeff), reps(1000): pxtmixed

Thanks for any advice on these questions, 
Jessica

*
*   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/