Dear Statalists.
In Roodman (2009) mentions that having too many instruments can weaken the Hansen test. Is there a specific ratio or threshold between the number of instruments and the number of groups (e.g., 20%, 50%, or 70%)?
I am currently running a System GMM (S-GMM) model with 544 instruments and 659 groups, but the p-value of the Hansen test (Hansen test of overid. restrictions) is only 0.145. Is this result reliable?
Thank you!
In Roodman (2009) mentions that having too many instruments can weaken the Hansen test. Is there a specific ratio or threshold between the number of instruments and the number of groups (e.g., 20%, 50%, or 70%)?
I am currently running a System GMM (S-GMM) model with 544 instruments and 659 groups, but the p-value of the Hansen test (Hansen test of overid. restrictions) is only 0.145. Is this result reliable?
Thank you!
Comment