i´m adjusting a temporal series into a ARIMA model to do some interventions analysis... After adjusting the segment to predict values to the future, i start to doubt which method is correct:

  1. The prediction of 66 values in one step, or

  2. The prediction of 1 value, and then this was included to the series to predict the next one (the idea was to reduce the error) until completing the 66.

My doubts rely on the fact that both methods gave me same results. So is this correct? Thanks.

  • 1
    $\begingroup$ A 66-steps-ahead forecast should not yield the same forecast as a 1-step-ahead forecast using a model that has 65 more historical data points. I suspect an error. Or are you updating your model using your forecasts, instead of a "rolling origin" approach where you successively add more observations? $\endgroup$ – Stephan Kolassa Oct 22 '15 at 14:53
  • $\begingroup$ You may have better luck getting a response if you edit your title to be more reflective of your problem. $\endgroup$ – Tchotchke Oct 22 '15 at 17:32
  • $\begingroup$ Stephan, i thank you for your comments. After several analyses, i found out that i was using the forecasted values to feed my model, i think that´s the reason of my results. I still doubting of using this approach (1-step-ahead forecast), because the main objective of my work is to analyse (quantify) the intervention effect, instead of building the best model to explain the observed values. Can you give me any comment about this? $\endgroup$ – ALFONSO Oct 26 '15 at 13:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.