specifies the sort order for the levels of classification variables. This order determines which parameters in the model correspond to each level in the data. The following table shows how to interpret values of the ORDER= keywords:
| Value of ORDER= | Levels Sorted By |
|---|
| FORMATTED | External formatted values, except for numeric variables that have no explicit format, which are sorted by their unformatted (internal) values. The sort order is machine-dependent. For numeric variables for which you have supplied no explicit format, the levels are ordered by their internal values. |
| FREQ | Descending frequency count (levels that have more observations come earlier in the order) |
| INTERNAL | Unformatted value. The sort order is machine-dependent. |
The ORDER= option does not apply to missing values, which are always ordered first.
For more information about sort order, see the chapter about the SORT procedure in Base SAS Procedures Guide and the discussion of BY-group processing in the "Grouping Data" section of SAS Programmers Guide: Essentials. By default, ORDER=FORMATTED.
specifies the parameterization method for the classification variable or variables. You can specify any of the keywords shown in the following table; design matrix columns are created from CLASS variables according to the corresponding coding schemes.
Table 2: Value of PARAM=
| Value of PARAM= | Coding |
|---|
| EFFECT | Effect coding. The REF= option in the CLASS statement determines the reference level. |
| GLM | Less-than-full-rank reference cell coding. This keyword can be used only as a global-option and is applied to all CLASS variables; all other individual variable parameterization specifications are ignored. The REF= option in the CLASS statement indirectly determines the reference level through the order of levels. |
ORDINAL |
THERMOMETER | Cumulative parameterization for an ordinal CLASS variable
|
POLYNOMIAL |
POLY | Polynomial coding. If the classification variable is numeric, then the ORDER= option in the CLASS statement is ignored, and the internal unformatted values are used. |
REFERENCE |
REF | Reference cell coding. The REF= option in the CLASS statement determines the reference level. |
| ORTHEFFECT | Orthogonalizes PARAM=EFFECT coding. The REF= option in the CLASS statement determines the reference level. |
ORTHORDINAL |
ORTHOTHERM | Orthogonalizes PARAM=ORDINAL coding
|
| ORTHPOLY | Orthogonalizes PARAM=POLYNOMIAL coding. If the classification variable is numeric, then the ORDER= option in the CLASS statement is ignored, and the internal unformatted values are used. |
| ORTHREF | Orthogonalizes PARAM=REFERENCE coding. The REF= option in the CLASS statement determines the reference level. |
All parameterizations are full rank, except for the GLM parameterization. If you specify a full rank parameterization for any CLASS variable, then every CLASS variable without a specified coding is given the EFFECT coding.
By default, PARAM=GLM. For more information about how parameterization of classification variables affects the construction and interpretation of model effects, see the section Specification and Parameterization of Model Effects.
specifies that design matrix columns that correspond to any effect that contains a split classification variable can be selected to enter or leave a model independently of the other design columns of that effect. This option applies to procedures that perform model selection.
Suppose that the variable temp has three levels (hot, warm, and cold), that the variable gender has two levels (M and F), and that the variables are used in a PROC SEVSELECT run as follows:
proc sevselect data=mylib.data;
loss y;
class temp gender / split;
scalemodel gender gender*temp;
run;
The two effects in the SCALEMODEL statement are split into eight independent effects. The effect "gender" is split into two effects that are labeled "gender_M" and "gender_F". The effect "gender*temp" is split into six effects that are labeled "gender_M*temp_hot", "gender_F*temp_hot", "gender_M*temp_warm", "gender_F*temp_warm", "gender_M*temp_cold", and "gender_F*temp_cold". The previous PROC SEVSELECT step is equivalent to the following:
proc sevselect data=mylib.data;
loss y;
scalemodel gender_M gender_F
gender_M*temp_hot gender_F*temp_hot
gender_M*temp_warm gender_F*temp_warm
gender_M*temp_cold gender_F*temp_cold;
run;
The SPLIT option can be used on individual classification variables. For example, consider the following PROC SEVSELECT step:
proc sevselect data=mylib.data;
loss y;
class temp(split) gender;
scalemodel gender gender*temp;
run;
In this case, the effect "gender" is not split and the effect "gender*temp" is split into three effects, which are labeled "gender*temp_hot", "gender*temp_warm", and "gender*temp_cold". Furthermore, each of these three split effects now has two parameters that correspond to the two levels of "gender." The PROC SEVSELECT step is equivalent to the following:
proc sevselect data=mylib.data;
loss y;
class gender;
scalemodel gender gender*temp_hot gender*temp_warm gender*temp_cold;
run;
<variable <(options)>> </ global-options>;