Parameter-Order Subcommands (ARIMA command)

P, D, Q, SP, SD, and SQ can be used as additions or alternatives to the MODEL subcommand to specify particular lags in the model and degrees of differencing for fitting single or nonsequential parameters. These subcommands are also useful for specifying a constrained model. The subcommands represent the following parameters:

P. Autoregressive order.

D. Order of differencing.

Q. Moving-average order.

SP. Seasonal autoregressive order.

SD. Order of seasonal differencing.

SQ. Seasonal moving-average order.

  • The specification on P, Q, SP, or SQ indicates which lags are to be fit and can be a single positive integer or a list of values in parentheses.
  • A single value n denotes lags 1 through n.
  • A single value in parentheses, for example (n), indicates that only lag n should be fit.
  • A list of values in parentheses (i, j, k) denotes lags i, j, and k only.
  • You can specify as many values in parentheses as you want.
  • D and SD indicate the degrees of differencing and can be specified only as single values, not value lists.
  • Specifications on P, D, Q, SP, SD, and SQ override specifications for the corresponding parameters on the MODEL subcommand.

Example 

ARIMA SALES
  /P=2
  /D=1.
ARIMA INCOME
  /MODEL=LOG NOCONSTANT
  /P=(2).
ARIMA VAR01
  /MODEL=(1,1,4)(1,1,4)
  /Q=(2,4)
  /SQ=(2,4).
ARIMA VAR02
  /MODEL=(1,1,0)(1,1,0)
  /Q=(2,4)
  /SQ=(2,4).
  • The first command fits a model with autoregressive parameters at lags 1 and 2 (P=2) and one degree of differencing (D=1) for the series SALES. This command is equivalent to: 
    ARIMA SALES
  /MODEL=(2,1,0).
  • In the second command, the series INCOME is log transformed and no constant term is estimated. There is one autoregressive parameter at lag 2, as indicated by P=(2).
  • The third command specifies a model with one autoregressive parameter, one degree of differencing, moving-average parameters at lags 2 and 4, one seasonal autoregressive parameter, one degree of seasonal differencing, and seasonal moving-average parameters at lags 2 and 4. The 4’s in the MODEL subcommand for moving average and seasonal moving average are ignored because of the Q and SQ subcommands.
  • The last command specifies the same model as the previous command. Even though the MODEL command specifies no nonseasonal or seasonal moving-average parameters, these parameters are estimated at lags 2 and 4 because of the Q and SQ specifications.