Today’s email question: I work within a government budget office and sometimes have to forecast fairly simple time series several quarters into the future. Auto.arima() works great and I often get something along the lines of: ARIMA(0,0,1)(1,1,0)[12] with drift as the lowest AICc. However, my boss (who does not use R) takes issue with low-order AR and MA because “you’re essentially using forecasted data to make your …