Using the over-identifcation in NLSYSTEM

[Kraus]

I have a system with two free parameters and six equations. The system thus is over-identified. When I run the code in RATS using NLSYSTEM, it tells me that it is done by 'Non-Linear System Estimation'. In my understanding, this simply means that the weighting matrix is an identity matrix. Now, if a want to use the over-identification and have an optimized weighting matrix (GMM), I simply add a constant as an instrument. Correct?

[TomDoan]

If the six equations are all moment conditions (that is, they theoretically have mean zero), then yes, you want

instruments constant
nlsystem(inst,other options) / list of moment FRML's.