The X12 Procedure

SEATS Decomposition

PROC X12 can decompose the B1 series by using the SEATS decomposition method described in Gómez and Maravall (1997a, 1997b). The SEATS decomposition method is planned for inclusion in the U.S. Census Bureau’s X13 program, which is not yet available for release.

The SEATS method requires the series to be extended with the same number of backcast and forecast observations. The number of observations backcast and forecast must meet the following minimum criteria:

  • The number of forecast and backcast observations must be at least twice the number of observations in a year, with a minimum of 8.

  • The number of forecast and backcast observations must be at least $2 \times (q + Q*s)$, where the ARIMA model used to extend the series is $(p d q)(P D Q)s$ in standard Box-Jenkins notation.

  • The number of forecast and backcast observations must be at least $p + d + q + (P + D + Q)*s$, where the ARIMA model used to extend the series is $(p d q)(P D Q)s$ in standard Box-Jenkins notation.

If you specify the SEATSDECOMP statement and the number of forecasts or backcasts (either the default number or the number you specify) is not sufficient for SEATS decomposition, then the number of forecasts or backcasts is increased to the minimum required.