I would like to seek your help in guiding me in backcasting those lost observations due to differencing to achieve stationarity of a time series. The ARIMA procedure does provide an option called "back=" in the forecast statement. However, it cannot help backcast any lost observations. Please advise.