Dear Stata-listers,
I have a model as followed:
(1) Yit = Xit,1*b1 + Xit,2*b2i + u1it
(2) b2i = Zit*d + u2it,
where i is a cross-sectional unit, t is time index, Xit and Yit is input and output,
respectively. Xit,1 is assumed to be exogenous and to affect Y in an autonomous way,
Zit is exogenous, u1it~N(0;sigma(u1)), u2it~N(0;sigma(u2)).
The idea is to allow the coefficients to vary with i.
Of course, a simple solution to the problem may be to interact Xit,2 and Zit:
Yit = Xit,1*b1 + Xit,2*b2 + Xit,2*Zit*b3 + u3it.
Than, the cross-section specific coefficient should be bi=b2+b3*Zit. Shortcoming of
that approach may consist, however, in introducing multi-collinearity between interacted
terms, if there are many of them.
Is there some multilevel approach in Stata? Does anybody know how to estimate that in
Stata? Is there a common procedure in Stata? Thanks for any suggestions.
Viktor
