Hello there. I am using a panel data of 110 obs for 8 years and I am interested in adopt the GMM model; in Stata, I use the commanda xtabond2; i have question related the specification of the model:
1. May I use eq (level) and eq (diff) specification both in IV equation as well as in GMM equatio?
2. I have many dummies. For example, I have spatial dummies that define the geographical pattern of Italy: Dummy for North, for Centre and for South; May I have decline one of these dummies in doing GMM regression in ordern to avoid the dummy variable trap? Or, using GMM, is that procedure done automatically? Or May do i have to drop one of these already mentioned dummies manually?
3. In doing GMM with 2 lags analysis using xtabond2 command, I found a problems of unit root in the coefficients (sum of lagged coefficients returns 1 almost); hence, what I have to do when a problem of unit root appear in lagged coefficients especially? May I have to transform my variables into logarithmic ones? What do you suggest to do when unit root problems arise in GMM dynamic model?

thanks to those who will take few minutes in providing a complete Answer kindly. In addition, i would thank in advance for your suggestions.
Kind Regards
Alessandra