0

I plan to forecast daily average temperature via (S)ARIMA model. I want to simulate the forecasting process many times in order to get a distribution of the temperature at a given time.

Does ARIMA include a stochastic component, hence I wouldn't get the same values every time?

Thanks!

user163494
  • 151
  • ARIMA includes an error term (and its lags) which is stochastic. But if you fit the same model to the same data, then each time you will get the same result (the stochastic component would be estimated to be the same regardless of how many times you try). – Richard Hardy Jun 22 '17 at 13:19
  • Okay. Actually I want to fit it once to the data and use the same model for prediction. But isn't there a white noise component in the model?

    Any idea how to get different values in the forecast?

     – user163494 Jun 22 '17 at 13:26
  • Bootstrap from the estimated white noise component either at the stage of forecasting or at the stage of fitting (after an initial fit that yields the white noise component). – Richard Hardy Jun 22 '17 at 13:36
  • Thank you very much. Can you be maybe more specific on how to conduct this step in R, as I dont think I understand it entirely.

    At which stage is it better to do so?

     – user163494 Jun 22 '17 at 13:42
  • Not at the moment, I cannot. Which stage depends on what you aim for. – Richard Hardy Jun 22 '17 at 13:47

0 Answers0