ARIMA transfer model is not computing forecasts

Technote (FAQ)


Question

I am using using the Forecasting module in SPSS Statistics to model a time series and forecast values for 6 months into the future. I have requested an ARIMA model with 2 predictors. The coefficients and fit statistics for the model, as printed in the Statistics Viewer, seem reasonable and their are no error messages. However, the estimated series values for the future time points (6 cases at the end of the file) are system missing. Why did the procedure fail to forecast the series for these time points?

Answer

If there is a predictor in the model, then there must be a value for the predictor in cases to be forecast.

If the forecast is into the future, so that the predictor values are not known, they would still need to be
estimated before the transfer function series is run. One work-around is to use an exponential smoothing model or an ARIMA model without a predictor to forecast the transfer function predictor(s) up to the point where you wish to forecast the dependent variable of interest. Suppose you eventually want to predict Y from X, forecasting to Jan 2015. You could run an exponential smoothing model on X, forecasting to Jan. 2015 and saving the predicted value. In your transfer function for Y, the predictor would now be the predicted value of X as saved by that exponential smoothing run.


Rate this page:

(0 users)Average rating

Document information


More support for:

SPSS Statistics

Software version:

Not Applicable

Operating system(s):

Platform Independent

Reference #:

1621609

Modified date:

2013-04-26

Translate my page

Machine Translation

Content navigation