FORECAST Subcommand (ARIMA command)

The FORECAST subcommand specifies the forecasting method to use. Available methods are:

EXACT. Unconditional least squares. The forecasts are unconditional least squares forecasts. They are also called finite memory forecasts. This is the default.

CLS. Conditional least squares using model constraint for initialization. The forecasts are computed by assuming that the unobserved past errors are zero and the unobserved past values of the response series are equal to the mean.

AUTOINIT. Conditional least squares using the beginning series values for initialization. The beginning series values are used to initialize the recursive conditional least squares forecasting algorithm.