Hi All:

I'm working on estimation with the Stochastic Frontier Analysis (fels) in the context of a panel data set and I've been having trouble with missing values in some spatial units. To handle this, I interpolated the variables I'm using in the model (nnipolate), and since some units persisted with missing values, I filtered the data, keeping only observations with information (keep if var!=.) and taking care that the panel is still balanced.

Is this a good approach? If not, what can you recommend?

Thank you!