st: gllamm multinomial logit

[kasakawa@ualberta.ca]

Dear all,

I am trying to run a two-level multinomial random-intercept model that  
includes level-2 covariates.

What I have is a dependent variable Y which takes 3 nominal values and  
X1 & X2 are level-1 control variables and Z1 & Z2 are level-2  
variables where the intercept is a function of Z1 and Z2.   AREA is a  
geographical identification variable.

I am wondering if anyone can evaluate the following gllamm syntax.  In  
particular, I am not quite sure if coefficients on Z1 and Z2 really  
represents effects of Z1 and Z2 on latent variable (intercept).

Any comments will be greatly appreciated.

-----
sort AREA X1 X2 Y Z1 Z2
gen patt=_n
expand 3
sprt patt
qui by patt: gen dep=_n
gen chosen=dep==Y
sort patt dep
tab dep, gen(a)

gen a2_X1=a2*X1
gen a3_X1=a3*X1
gen a2_X2=a2*X2
gen a3_X2=a3*X2
gen a2_Z1=a2*Z1
gen a3_Z1=a3*Z1
gen a2_Z2=a2*Z2
gen a3_Z2=a3*Z2

eq a2: a2
eq a3: a3

gllamm dep a2_X1 a2_X2 a2_Z1 a2_Z2  a3_X1 a3_X2 a3_Z1 a3_Z2, i(AREA)  
nrf(2) eq (a2 a3) expand(patt chosen o) lin(mlogit) fam(binom) nip(4)  
adapt trace
---------

end of syntax


*
*   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/