A parameterization of a CLASS variable is the representation of the variable by a set of coded variables (sometimes called indicator or dummy variables). The most common parameterization is that used in PROC GLM in which a CLASS variable with k levels is represented in the design matrix by a set of k 0,1-coded dummy variables. This is known as GLM parameterization. Other parameterizations code the variables differently and, in the context of a fitted model, result in equivalent models (same fit statistics) with different interpretations of the parameters. In most procedures, a variable is declared a CLASS variable by specifying it in the CLASS statement. However, CLASS variables are declared differently in some procedures or are the default. For details on the different parameterizations of CLASS variables, see "Parameterization of Model Effects" in the Shared Concepts and Topics chapter of the SAS/STAT^® Users Guide. For more information about each procedure's CLASS parameterization, see the procedure's documentation.

Note that many procedures such as GLM, MIXED, GLIMMIX, LIFEREG and others do not allow different parameterizations of CLASS variables. These procedures offer only the GLM parameterization. Some of these procedures only use the last level (when the levels are sorted) of a CLASS variable as the reference level. Beginning in SAS^® 9.3 TS1M2, the GLM, MIXED, GLIMMIX, PROBIT, and ORTHOREG procedures allow reference levels to be set using the REF= option in their CLASS statements. This note describes how to select a different reference level for a CLASS predictor in those procedures which do not provide a REF= option.

Prior to SAS 8.0, the TRANSREG procedure was the only procedure that supported different parameterizations of CLASS variables. In later releases, other procedures acquired additional CLASS variable parameterizations including various full-rank parameterizations. The following table lists the SAS/STAT^® and SAS/ETS^® procedures which allow multiple CLASS parameterizations, how CLASS variables are declared, what the default parameterization is, the SAS release in which they acquired that capability, and how the parameterization is selected. Note that many additional modeling procedures available in SAS^® Viya^® also provide multiple parameterizations using the PARAM= option in the CLASS statement.

Operating System and Release Information