Forecast Server Scenario analysis - Use time series models to predict exogenous variables incorporated as independent in a model

[andreas_zaras]

Hello,

I have a dependent variable and an independent variable that is continous. I am using the automatic model generation facility in Forecast Server and the software ends up with a dynamic regression model. In the scenario analysis view i can see that the software has predicted future values for the independent variable that seem to be like a linear trend or a moving average of the past values of the variable (independednt). Then the dependent variable is forecasted based on this trend or moving average of the independent variable. Is there a way that i can tell Forecast Server to use a time series model (like ARIMA or smoothing) to predict the independent variable and based on this to predict the dependent through the regression model? What i am doing until, now is first to forecast the independent variable using time series models (ARIMA or smoothing) and then add this values to the end of the data set with the dependent variabe's values as missing.Then i fit the (regression) model and i create a scenario. The forecasted values for the independent variabe are now the forecasts from the time series model (e.g. ARIMA or smoothing) and the forecasted values of the dependent variable are produced by the regression model.

Thnaks in advance,

Andreas

[udo_sas · Accepted solution]

Hello Andreas -

If you don't provide future values of independent variables, they are automatically forecast using one of the following smoothing models:

• simple
• double
• linear
• damped trend
• seasonal
• Winters method (additive and multiplicative)

On the other hand, if you provide future values of your independent variables (similar to what you describe), these values will be taken into account.

Thanks,

Udo

[andreas_zaras]

Thanks very much Udo!