The STATESPACE Procedure

Overview: STATESPACE Procedure

The STATESPACE procedure has largely been superseded by the newer SSM procedure . PROC SSM fits and forecasts very general linear state space models. It supports irregularly spaced time series and replicated longitudinal data, in addition to supporting regular fixed-period time series. The SSM procedure also provides a powerful expressive language for specifying state space models, and allows programming statements to define model elements through user-written functions of unlimited complexity. The SSM procedure also provides more modern estimation, filtering, and forecasting algorithms than the older STATESPACE procedure. See Chapter 27: The SSM Procedure, for information about PROC SSM.

Although the SSM procedure should be preferred to the STATESPACE procedure for most state space modeling applications, the STATESPACE procedure should be considered if you wish to perform automated multivariate forecasting using a state space model selected through the modeling strategy proposed by Akaike (1976). This strategy employs an initial sequence of unrestricted vector autoregressive (VAR) models, selection of an initial VAR model using Akaike’s information criterion (AIC), followed by a canonical correlation analysis for the automatic identification of the state space model to use to forecast the vector of time series.

The operation of the STATESPACE procedure and the form of state space model it supports are described in the following.

The STATESPACE procedure uses the state space model to analyze and forecast multivariate time series. The STATESPACE procedure is appropriate for jointly forecasting several related time series that have dynamic interactions. By taking into account the autocorrelations among all the variables in a set, it is possible that the STATESPACE procedure may give better forecasts than methods that model each series separately.

By default, the STATESPACE procedure automatically selects a state space model appropriate for the time series, making the procedure a good tool for automatic forecasting of multivariate time series. Alternatively, you can specify the state space model by giving the form of the state vector and the state transition and innovation matrices.

The methods used by the STATESPACE procedure assume that the time series are jointly stationary. Nonstationary series must be made stationary by some preliminary transformation, usually by differencing. The STATESPACE procedure enables you to specify differencing of the input data. When differencing is specified, the STATESPACE procedure automatically integrates forecasts of the differenced series to produce forecasts of the original series.