Non-Linear GMM Weighting Matrix Issues
Posted: Wed Apr 01, 2015 3:30 pm
I am trying to estimate a non-linear Cobb-Douglas production function (my workfile is attached) of the following form( I am using eviews 9):
y=c(1) * K^c(2) * L^c(3)
I am trying to estimate this GMM using a user defined matrix. In theory for estimating Optimal GMM you follow a two-step process:
1) Assume the Weighting Matrix is S^-1 and we estimate GMM by setting the weighting matrix equal to the identity matrix. Then use the estimated s from the first equation and use that estimated S_hat^-1 as the weighting matrix
However when I run this estimation and select a user defined matrix (i in this workfile is the identity matrix) I get that there is a singluar matrix.
In this estimation I do not have any IV's so I am using the variables themselves as the IV's.
Perhaps this is the issue? Is there a way to code the objective function for GMM and maximize it if this does not work?
The equation nonlineargmm is the estimation using a a different weighting matrix (which does not give me a nonsingluar result)
Thanks,
Tim
y=c(1) * K^c(2) * L^c(3)
I am trying to estimate this GMM using a user defined matrix. In theory for estimating Optimal GMM you follow a two-step process:
1) Assume the Weighting Matrix is S^-1 and we estimate GMM by setting the weighting matrix equal to the identity matrix. Then use the estimated s from the first equation and use that estimated S_hat^-1 as the weighting matrix
However when I run this estimation and select a user defined matrix (i in this workfile is the identity matrix) I get that there is a singluar matrix.
In this estimation I do not have any IV's so I am using the variables themselves as the IV's.
Perhaps this is the issue? Is there a way to code the objective function for GMM and maximize it if this does not work?
The equation nonlineargmm is the estimation using a a different weighting matrix (which does not give me a nonsingluar result)
Thanks,
Tim