# Aggregating forcast to higher spatial level

I am working on a project doing out of sample estimates of wheat yields at the village level in STATA. I am using a short panel 3 years and a sample of ~3000 villages. We are estimating a RE model and then predicting out of sample to the remaining villages not included in our sample for each of the 3 years. We estimate following model:

 Yit = a + b1x1 + .. + Uit + Eit


I am using Yhat to estimate out of sample yield estimates and Yhat+-1.96xstdp for 95% confidence bounds. Where stdp is the standard error of the fitted values.

I was recently just asked to make estimates at a higher level village > district and would prefer to aggregate my village level estimates up to the district rather than re-estimating the model at the district level.

I am assuming that this is possible but not sure how to implement it (aggregating Yhat and stdp). Please keep in mind that I am more of an applied econometrician than a theoretical one. If your answer could be specific to STATA that would also be appreciated!