You are not logged in. You can browse but not post. Login or Register by clicking 'Login or Register' at the top-right of this page. For more information on Statalist, see the FAQ.
I suspect Anne means checking the relevancy requirement -- that is, determining whether the IVs are sufficiently strong. Joao's point generally stands, but there are things one can do. For example, if you use the control function version of the estimator, the first step regression is exactly the same as the first step in linear 2SLS. Thus, as a rough guide, you can perform the same Stock-Yogo and other tests for weak instruments. In Stata 16 you can juse the "estat first" option after ivregress. In earlier versions of Stata, ivreg2 can be used. That you are estimating a linear model is not important for getting the Stock-Yogo and other diagnostics. Actually, I think a linear model estimated by 2SLS is not a bad idea, anyway. You can see whether the directions of effects line up with ivpoisson.
If you use the GMM estimator in ivpoisson things are more difficult. I'm a bit behind in this literature, but I don't know that we have good ways of establishing sufficient relevance in models nonlinear in the parameters.
If you only have more IVs than you need, you can also test the overidentification assumption as a crude check on exogeneity of the IVs.
Comment