Suppose I have an ARIMA model with $n$ step ahead forecasts and forecast intervals. If I want to sample from those forecasts:

  • One way to look at is: take one observation from the prediction distribution in each step, then you have one path. Repeat this process.

  • But when you choose your first observation from the first prediction distribution, the second observation is dependent on the first one, isn't it?

Any insights?

  • 1
    $\begingroup$ Yes, you are right. And of course we want this dependence. If your time series were independent, there would not be a point to time series modeling. Does this answer your question, or could you clarify if something is still unclear to you? $\endgroup$ – Stephan Kolassa Nov 1 '17 at 7:28
  • $\begingroup$ What is the exact question here? Also note that ARIMA is an acronym, so we do not use formula formatting for it, we just use capital letters. $\endgroup$ – Richard Hardy Nov 1 '17 at 11:03
  • $\begingroup$ @StephanKolassa, yes it does. It seems that simulate function in forecast package in R does the job for sampling from the forecasts. This link is useful. $\endgroup$ – soheil Nov 1 '17 at 14:15

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.