How does PROC ARIMA score/forecast a new time series?

[davidchenus]

The PROC ARIMA goes through 3 steps of IDENTIFY, ESTIMATE and FORECAST. They are all related to the input time series. Assume the ARIMA model will use additional input variables to forecast a target variable. Now I would like to apply the model results to a new time series. The new time series contains those input variables (with values) and the target variable (with values) in the historical part and only the input variables (with values) in the future part.  How can I score/forecast the target variable in the future part of the new time series ?

[Puwang · Accepted solution]

The short answer to your question: there is no "score" functionality in PROC ARIMA. You will always need to re-estimate based on the history data.

That said, you can use the IDENTIFY statement to determine the p,d,q based on one time series. Then, use this pre-determined p,d,q to ESTIMATE and FORECAST a new set of time series. Hope this answer your question.