Calling MATA optimization routine through a STATA program

[Amaresh Tiwari]

Dear Experts,

I have written the following codes for GMM estimation. When I call GMM_OPTI(init) outside the STATA program, "struc_esti", using ---mata: GMM_OPTI(st_matrix("init")) --- I get my optimal estimates of "theta."

But when I call GMM_OPTI(init) through "struc_esti," I get the error message: *: 3200 conformability error. And when I read the elements of X and L_X element by element in CRIT_FN(todo,theta,crit,g,H) and call GMM_OPTI(init) through "struc_esti," struc_esti simply returns the initial values without performing the optimization.

I need to call GMM_OPTI(init) through "struc_esti" as need to do bootstrapping.

I wounder what I am missing. Any help would be most welcomed.

Best
Amaresh


#delimit ;
global omg_endo_poly "C, L1_OMG, L1_OMG:^2, L1_OMG:^3,
L1_OG, L1_CA, L1_CA:*L1_OG, L1_CA:^2, (L1_CA:^2):*L1_OG,
L1_OMG:*L1_OG, L1_OMG:*L1_CA, L1_OMG:*L1_CA:*L1_OG, L1_OMG:*(L1_CA:^2), L1_OMG:*(L1_CA:^2):*L1_OG,
(L1_OMG:^2):*L1_OG, (L1_OMG:^2):*L1_CA, (L1_OMG:^2):*L1_CA:*L1_OG, (L1_OMG:^2):*(L1_CA:^2), (L1_OMG:^2):*(L1_CA:^2):*L1_OG,
(L1_OMG:^3):*L1_OG, (L1_OMG:^3):*L1_CA, (L1_OMG:^3):*L1_CA:*L1_OG, (L1_OMG:^3):*(L1_CA:^2), (L1_OMG:^3):*(L1_CA:^2):*L1_OG";
#delimit cr

*Defining mata routines
capture mata mata drop CRIT_FN() // For evaluating the criterion function.
capture mata mata drop GMM_OPTI() // For minimizing the criterions function with respect to the structural parameters.
* MATA programms ************************************************** *******
mata:

void CRIT_FN(todo,theta,crit,g,H)
{
stru_vars = st_local("Struc_names")
st_view(X =., ., tokens(stru_vars) )
l_stru_vars = st_local("l_Struc_names")
st_view(L_X =., ., tokens(l_stru_vars) )

PHI=st_data(.,("phi"))
L_PHI=st_data(.,("l_phi"))

L_CA=st_data(.,("l_cua"))
L_OG=st_data(.,("l_dorg"))

inst_vars = st_local("inst_vars")
st_view(Z =., ., tokens(inst_vars) )

OMG=PHI-(X)*theta'
L_OMG=L_PHI-(L_X)*theta'
L_OMG_pol=(${omg_endo_poly})

beta_prod=invsym(L_OMG_pol'L_OMG_pol)*L_OMG_pol'OM G
XI=OMG-L1_OMG_pol*beta_prod

W = invsym(Z'Z)/(rows(Z))
crit=(Z'XI)'*W*(Z'XI)

}

void GMM_OPTI(init)
{
S=optimize_init()
optimize_init_evaluator(S, &CRIT_FN())
optimize_init_evaluatortype(S, "d0")
optimize_init_technique(S, "nm")
optimize_init_nmsimplexdeltas(S, 0.0001)
optimize_init_which(S, "min")
optimize_init_params(S, init)
p=optimize(S)
p
st_matrix("theta",p)
}
end

capture program drop struc_esti
program define struc_esti, rclass

*GMM_OPTI and CRIT_FN() are two MATA routines for estimating the structural parameters by the method of GMM.
*GMM_OPTI minimizes the criterion function evaluated in CRIT_FN() with respect to the structural parameters and returns the
*optimal structural parameters.

mata: GMM_OPTI(st_matrix("init"))

return scalar Dummy_OCC =theta[1,1]
return scalar Dummy_OCC =theta[1,2]
return scalar Dummy_REG =theta[1,3]

return scalar Labor_NON =theta[1,4]
return scalar Labor_OCC =theta[1,5]
return scalar Labor_REG =theta[1,6]

return scalar Capit_NON =theta[1,7]
return scalar Capit_OCC =theta[1,8]
return scalar Capit_REG =theta[1,9]

return scalar Harjumaa_NON=theta[1,10]
return scalar Harjumaa_OCC=theta[1,11]
return scalar Harjumaa_REG=theta[1,12]

end