The HPFENGINE Procedure

(View the complete code for this example.)

The HPFDIAGNOSE procedure is often used in conjunction with the HPFENGINE procedure. This example demonstrates the most basic interaction between the two. In the following statements, model specifications are created by the HPFDIAGNOSE procedure and are those specifications are then fit to the data by using the HPFENGINE procedure.

proc hpfdiagnose data=sashelp.air
                 repository=work.repository
                 outest=est;
   id date interval=month;
   forecast air;
run;
proc hpfengine data=sashelp.air
               inest=est outest=outest
               repository=work.repository
               print=(select estimates summary)
               plot=forecasts;
   id date interval=month;
   forecast air;
run;

The HPFENGINE procedure output is shown in Output 5.5.1. Forecasts are shown in Output 5.5.2.

It is also possible to compare the performance of model specifications that you create with those automatically generated by the HPFDIAGNOSE procedure.

In the following statements, two model specifications are created, together with a selection list to reference those specifications.

proc hpfarimaspec repository=work.repository name=myarima;
   forecast symbol=sale transform=log q=(1 12) diflist=(1 12) noint;
run;

proc hpfucmspec repository=work.repository name=myucm;
   forecast symbol=sale transform=log;
   irregular;
   level;
   season length=12;
run;

proc hpfselect repository=work.repository name=select;
   spec myarima myucm;
run;

The model selection list is passed to the HPFDIAGNOSE procedure by using the INSELECTNAME= option as follows. The new selection list ultimately created by the HPFDIAGNOSE procedure includes all model specifications in the INSELECTNAME= selection list together with automatically generated model specifications.

proc hpfdiagnose data=sashelp.air
                 inselectname=select
                 repository=work.repository
                 outest=est
                 criterion=mape;
   id date interval=month;
   forecast air;
run;

The HPFENGINE procedure selects the best-fitting model as follows. Selection results are shown in Output 5.5.3, parameter estimates of the selected model are shown in Output 5.5.4, and the forecast summary is shown in Output 5.5.5.

proc hpfengine data=sashelp.air
               inest=est
               repository=work.repository
               print=(select estimates summary)
               plot=forecasts;
   id date interval=month;
   forecast air;
run;