Hi everyone,

I am estimating a regression with panel data for the model WfpPcLog ~ GdpPcLog + Gdp2PcLog + Gdp3PcLog + HarvestedAreaPc + LivestockPc (water footprint as dependent variable and linear, quaratic and cubic GDP as independent variables plus others two covariates). I need to estimate the coefficients of the model with CCEMG and AMG estimators, but I'd like to weight the final coefficients for population, so that if a country has got, for example, 50% of the world population (I can use invariant weights in time), its coefficient is counted for 0.50, if a country has got the 5% the coefficient counts for 0.05 and so on.

I know that the XTMGcommand in stata, for both CCEMG and AMG, at first estimates N group-specific ordinary least-squares (OLS) regressions for each country and secondly it caluclates the final coefficient of the model as the average of the estimated coefficients across groups.

I was wondering if I just can use a specific vector of weights to calculate a weighted average of these estimated coefficients across groups, where more population the country has, more important will be his coefficient in the average. And, if it is possible, I don't understand how I could calculate the new standard error to test the significance of the new weighted coefficient found.

Just ask me if you need more information and thanks all