Getting Started: HPFARIMASPEC Procedure

The following example shows how to create an ARIMA model specification file. In this example the specification for an Airline model with one input is created.

   proc hpfarimaspec repository=work.arima
                     name=Airline1
                     label="Airline model with one input";
      forecast symbol=Y q=(1)(12) dif=(1, 12) noint
               transform=log;
      input symbol=X dif=(1, 12);
      estimate method=ml;
   run;

The options in the PROC HPFARIMASPEC statement are used to specify the location of the specification file that will be output. Here the REPOSITORY= option specifies that the output file be placed in the catalog SASUSER.ARIMA, the NAME= option specifies that the name of the file be Airline1.xml, and the LABEL= option specifies a label for this catalog member. The other statements in the procedure specify the ARIMA model and the options used to control the parameter estimation process for the model. The model specification begins with the FORECAST statement that specifies the following:

• transformation, such as or Box-Cox, and the differencing orders associated with the variable that is to be forecast

• autoregressive and moving average polynomials

• presence or absence of the constant in the model

According to the FORECAST statement, the model contains no constant term and has a two-factor moving average polynomial of orders 1 and 12. The forecast variable is transformed and differenced with differencing orders 1 and 12. The SYMBOL= option in the FORECAST statement can be used to provide a convenient name for the forecast variable. This name is only a placeholder, and a proper data variable is associated with this name when this model specification is used in actual data analysis.

Next, the INPUT statement provides the transfer function specification associated with the input variable in the model. In the INPUT statement you can specify the following:

• transformation, such as or Box-Cox, and the lagging and differencing orders associated with the input variable

• numerator and denominator polynomials associated with the transfer function input

In this case the input variable is differenced with differencing orders 1 and 12, and it enters the model as a simple regressor. Here again the SYMBOL= option can be used to supply a convenient name for the input variable. If a model contains multiple input variables, then each input variable has to be specified with a separate INPUT statement.

Lastly, the ESTIMATE statement specifies that the model be estimated using the ML method of estimation.