The QTRSELECT Procedure
MODEL Statement
MODEL dependent=<effects> / <model-options>;
The MODEL statement names the dependent variable and the explanatory effects, including covariates, main effects, interactions, and nested effects. If you omit the explanatory effects, the procedure fits an intercept-only model.
After the keyword MODEL, the dependent (response) variable is specified, followed by an equal sign. The explanatory effects follow the equal sign.
For information about constructing the model effects, see the section Specification and Parameterization of Model Effects in Chapter 2, Shared Concepts.
The model-options control other aspects of model formation and inference. Table 5 summarizes these options.
Table 5: MODEL Statement Options
| Option | Description |
|---|---|
| Model Options | |
| CLB | Requests confidence limits |
| INCLUDE= | Includes effects in all models for model selection |
| INFORMATIVE | Models missing values by using extra indicator variables |
| NOINT | Suppresses the intercept |
| QUANTILES | Specifies quantile levels for the quantile regression |
| START= | Includes effects in the initial model for model selection |
| STB | Displays standardized estimates |
Model Options
You can specify the following options in the MODEL statement after a slash (/):
- CLB
requests the
% upper and lower confidence limits for the parameter estimates. By default, the 95% limits are computed; you can use the ALPHA= option in the PROC QTRSELECT statement to change the 
level. -
INCLUDE=n
INCLUDE=single-effect
INCLUDE=(effects) forces effects to be included in all models. If you specify INCLUDE=n, then the first n effects listed in the MODEL statement are included in all models. If you specify INCLUDE=single-effect or INCLUDE=(single-effect), then the specified effects are forced into all models.
- INFORMATIVE
models missing values by using extra model effects. These effects consist of dummy variables that take the value 1 when the value of a continuous model variable involved in the effect is missing, and take the value 0 otherwise. The missing value in the original model effect is replaced by the average value of the effect for the nonmissing values. For continuous-by-class effects, such as A*x, where A is a classification variable and x is a continuous variable, informative missingness creates multiple dummy columns and substitutes the effect mean of x that corresponds to the respective level of A. Missing values for classification variables are treated as valid levels. For more information about informative missingness, see the section Informative Missingness in Chapter 2, Shared Concepts.
- NOINT
suppresses the intercept term that is otherwise included in the model.
-
QUANTILES<(quantile-level-options)><=number-list>
QUANTILE<(quantile-level-options)><=number-list>
Q<(quantile-level-options)><=number-list> -
specifies the quantile levels for the quantile regression. You can request any values of quantile levels in
by specifying a number-list. By default, the QTRSELECT procedure uses QUANTILES=0.5, which corresponds to median regression. You can also specify the following quantile-level-options:
-
NTAU=n
NQ=n -
specifies the following n quantile levels for the quantile regression:
If you specify both a QUANTILES=number-list and NTAU=n, the QTRSELECT procedure uses all the specified quantile levels.
- SORT
sorts all the specified quantile levels in ascending order. This option affects the order of the prediction variables in the SAS DATA step code that is created by the CODE statement, and the order of the appropriate keyword variables in the SAS data set that is created by the OUTPUT statement.
-
NTAU=n
-
START=n
START=single-effect
START=(effects) begins the selection process in the FORWARD and STEPWISE selection methods from the initial model that you designate. If you specify START=n, then the starting model consists of the first n effects listed in the MODEL statement. If you specify START=single-effect or START=(single-effect), then the starting model consists of these specified effects.
- STB
-
produces standardized regression coefficients. A standardized regression coefficient is computed by dividing a parameter estimate by the ratio of the sample standard deviation of the dependent variable to the sample standard deviation of the regressor. If you use the INCLUDE= option to force some effects to be in the model, then the QTRSELECT procedure computes the sample standard deviation against all the effects that are forced in, as follows. Let
denote the design submatrix of 
regressors that consists of all the effects that are forced in, and let 
denote the dependent variable or any regressor. Then the sample standard deviation of is computed as 