Relationship between Instrumental variable regression and structural equation modelling (SEM)

[Araceli R]

Dear all

I would like to know the relationship between instrumental variable regression and structural equation modelling (SEM).
I need to transform a SEM model into iv regression model to endogeneity purpose.

Thank you very much in advance.

Araceli

[wbuchanan]

Your question isn't exactly clear. Not all structural equation models include analogs for an excluded instrument in an instrumental variable estimator. If it is a "full" structural equation model (I.e., has measurement and structural components), there wouldn't be an analogous way to translate the measurement model into an IVE model. Perhaps you could share the model you are attempting to refit and/or explain why there is a need to refit the same model perhaps using a different estimator?

[Araceli R]

Dear wbuchanan

Thank you for your interest.

I am trying to refit the model using iv regression because I want to perform the endogeneity test Durbin Watson Hausman.
Nevertheless I am not sure if that make sense.

[Richard Williams]

Here is a simple example, adapted from the ivregress help:

webuse hsng2, clear
ivregress liml rent pcturban (hsngval = faminc)
sem (rent <- pcturban hsngval) (hsngval <- faminc pcturban), cov(e.rent*e.hsngval) nolog

Basically I regressed hsngval on its instruments and then let its residual correlate with rent's residual. To me this setup is a little more intuitive than the usual ivregress setup.

[Araceli R]

Dear Richard

Thank you very much for your interest.
Sorry I couldn´t understand well your answer because I am not very familiar with STATA commands.
Would you mind to explain it more extensively?

Thank you in advance

Araceli

[Richard Williams]

Araceli, do you already have a SEM model you have estimated, or what? If you showed us what you already have it might be easier to advise you from there.

Also, if you run the code I posted above it may be easier for you to understand what the commands are doing.