spivregress first stage

[Dario Maimone Ansaldo Patti]

Dear all,

I have a question which I am unable to solve. When stata estimates a spatial model with an endogenous regressor, which model does it estimate at the first stage?

I tried to study spivregress.ado, which redirect to other functions, specifically to a Mata one, which is not accessible (I guess for proprietary reasons).

I tried to contact the stata technical support, but I did not get any answer, because they said that this is a theoretical question and not a technical one. They suggested to read the user manual, which in my view is not clear at all.

Therefore, I am left with my original doubt: which model is estimated at the first stage? Is it a standard linear regression model, estimated via ols?

Any feedback you may want to provide is highly appreciated.

Dario

[Andrew Musau]

You previously asked this question, and I provided a link to the paper that describes the implementation of the generalized spatial two-stage least squares (GS2SLS) estimator:
https://www.statalist.org/forums/for...on-first-stage.

I'm not sure why you want to estimate this model in steps when spivregress—which performs proper spatial IV estimation—already exists.

[Dario Maimone Ansaldo Patti]

Andrew Musau I apologize for cross-posting. The reason why I want to estimate the model in steps is that spivregress does not produce any diagnostic tests for instance to check how strong are instruments. Since I am going to start with a model with some instruments and then I add more, I would like to test if the augmented set of instruments is still strong or not