Hello,

I've noticed that the fixed effects results from plm are sensitive to how the data is ordered.

The fixed-effects regression provides the expected results when the data is sorted by unit-time (unit first and then time), but provides different (wrong) coefficients when the data is sorted by time-unit (time first then unit).

I think it would make a great package even better if the results were not sensitive to how the data is sorted, or if at the very least there is an error message that pops out alerting researchers that their data is not sorted the right way if the data is not pre-arranged as unit-year.

Thanks!