07-25-2012 03:02 PM
I have a panel data (unbalance, not dynamic data), and I want to use fixed effect to control within individual effects.
The model have simultaneous equations and I want to use GMM to estimate (PROC MODEL with GMM option)
Since there are way more individuals (100+) than periods (several), dummy variable means sacrifice a lot of degree of freedom. I wonder if mean deviation method would give the same results for GMM, compared with the dummy variable method? I am asking because OLS estimates with fixed effects dummy variables will give the same estimates as the mean deviation method. But for comparison of the two methods for GMM estimation method, I am not sure. Since with GMM estimation method, the instruments used will affect the parameter estimates, hence the comparison may be more complicated.