0

If I have Time series from 1960 to 2018 and i want to forecast data befor 1960, How I can do it in R? enter image description here

duaa
  • 93
  • 1
  • 6
  • 1
    I'm mot saying this is the best way, but how about reversing the order of the time series and forecasting into the future of the reversed time series? – Mark L. Stone Apr 03 '18 at 16:58
  • 2
    There might be some logical cut-off point for the past. Like the `AirPassengers` dataset in R, it would be meaningless to predict before 1903. I say this because it seems like you data set is close to 0 around the beginning. Is there a true 0 point for your variable? – Jay Schyler Raadt Apr 03 '18 at 17:17
  • Ok, if my data is AirPassengers dataset in R,how to predict before 1903? – duaa Apr 03 '18 at 17:59
  • You wouldn't predict air passengers before 1903 because it is impossible to have them before 1903. – RAND Apr 03 '18 at 19:50
  • 1
    https://stats.stackexchange.com/questions/190807/regression-on-default-data-and-backward-extrapolation/190822#190822 suggests that you do exactly what Mark l Stone stated. – IrishStat Apr 03 '18 at 20:05
  • 1
    For some code showing how to do it in R, see https://otexts.org/fpp2/backcasting.html – Rob Hyndman Apr 03 '18 at 22:16
  • @IrishStat Great minds thing alike. Although I make no claims to being the first one to think of reversing the time series to, as you say, backcast. – Mark L. Stone Apr 04 '18 at 15:21

0 Answers0