The ARIMA Procedure |

Naming of Model Parameters |

In the table of parameter estimates produced by the ESTIMATE statement, model parameters are referred to by using the naming convention described in this section.

The parameters in the noise part of the model are named as AR*i,j* or MA*i,j*, where AR refers to autoregressive parameters and MA to moving-average parameters. The subscript *i* refers to the particular polynomial factor, and the subscript *j* refers to the *j*th term within the *i*th factor. These terms are sorted in order of increasing lag within factors, so the subscript *j* refers to the *j*th term after sorting.

When inputs are used in the model, the parameters of each transfer function are named NUM*i,j* and DEN*i,j*. The *j*th term in the *i*th factor of a numerator polynomial is named NUM*i,j*. The *j*th term in the *i*th factor of a denominator polynomial is named DEN*i,j*.

This naming process is repeated for each input variable, so if there are multiple inputs, parameters in transfer functions for different input series have the same name. The table of parameter estimates shows in the "Variable" column the input with which each parameter is associated. The parameter name shown in the "Parameter" column and the input variable name shown in the "Variable" column must be combined to fully identify transfer function parameters.

The lag 0 parameter in the first numerator factor for the first input variable is named NUM1. For subsequent input variables, the lag 0 parameter in the first numerator factor is named NUM*k*, where *k* is the position of the input variable in the INPUT= option list. If the ALTPARM option is specified, the NUM*k* parameter is replaced by an overall scale parameter named SCALE*k*.

For the mean and noise process parameters, the response series name is shown in the "Variable" column. The lag and shift for each parameter are also shown in the table of parameter estimates when inputs are used.

Copyright © 2008 by SAS Institute Inc., Cary, NC, USA. All rights reserved.