specification
y_[it] = x_[it]*a_[it]+ w_[it]*b_[it]+ e_[it]
where i=cross sectional units (Banks) and t=time(Quarter). And
x_[it] are the co-variates and the error term e_[it] is assumed
to follow a Gaussian distribution. I have 45 time periods, 21
cross-sectional units and about 669 observations since this is an
unbalanced panel.
My first question: Is it correct, that in using -gllamm- to estimate
the above model I should do the following:
Step 1-Define an equation "slope" for co-variates that vary over
cross-sections (banks) and time (quarters)
Step 2-Then define another equation "slope2" for the co-variates
that vary only over time
eq slope: x1 x2 x3
eq slope2: x4 x5
gllamm y x1 x2 x3 x4 x5, i(Quarter Bank) nrf(1,1) nip(7,7) adapt
eqs(slope slope2) robust cluster(Bank) trace dots
On doing this however I fail to get any kind of convergence for the
maximum likelihood. Is there any suggestion on how I should go about this?
The second question: if -gllamm- is not appropriate for the above
model, is there a STATA command/ routine or some kind of work around that
would enable me estimate random coefficients that vary over time and
cross sections for unbalanced Panel data?
Many thanks,
Twinemanzi Tumubweinee (Twine)
PhD Student
School of Economics, Political and Policy Sciences
University of Texas at Dallas
Tel: 972-883-4913
[email protected]
*
* 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/
