The PLM Procedure

FILTER Statement

FILTER expression;

The FILTER statement enables you to filter the results of the PLM procedure, specifically the contents of ODS tables and the output data sets. There can be at most one FILTER statement per PROC PLM run, and the filter is applied to all BY groups and to all queries generated through WHERE expressions.

A filter expression follows the same pattern as a where-expression in the WHERE statement. The expressions consist of operands and operators. For more information about specifying where-expressions, see the WHERE statement for the PLM procedure and SAS Language Reference: Concepts.

Valid keywords for the formation of operands in the FILTER statement are shown in Table 75.6.

Table 75.6: Keywords for Filtering Results

Keyword	Description
Prob	Regular (unadjusted) p-values from t, F, or chi-square tests
ProbChi	Regular (unadjusted) p-values from chi-square tests
ProbF	Regular (unadjusted) p-values from F tests
ProbT	Regular (unadjusted) p-values from t tests
AdjP	Adjusted p-values
Estimate	Results displayed in "Estimates" column of ODS tables
Pred	Predicted values in SCORE output data sets
Resid	Residuals in SCORE output data sets.
Std	Standard errors in ODS tables and in SCORE results
Mu	Results displayed in the "Mean" column of ODS tables (this column is typically produced by the ILINK option)
tValue	The value of the usual t statistic
FValue	The value of the usual F statistic
Chisq	The value of the chi-square statistic
testStat	The value of the test statistic (a generic keyword for the 'tValue', 'FValue', and 'Chisq' tokens)
Lower	The lower confidence limit displayed in ODS tables
Upper	The upper confidence limit displayed in ODS tables
AdjLower	The adjusted lower confidence limit displayed in ODS tables
AdjUpper	The adjusted upper confidence limit displayed in ODS tables
LowerMu	The lower confidence limit for the mean displayed in ODS tables
UpperMu	The upper confidence limit for the mean displayed in ODS tables
AdjLowerMu	The adjusted lower confidence limit for the mean displayed in ODS tables
AdjUpperMu	The adjusted upper confidence limit for the mean displayed in ODS tables

When you write filtering expressions, be advised that filtering variables that are not used in the results are typically set to missing values. For example, the following statements select all results (filter nothing) because no adjusted p-values are computed:

proc plm restore=MyStore;
   lsmeans a / diff;
   filter adjp < 0.05;
run;

If the adjusted p-values are set to missing values, the condition adjp < 0.05 is true in each case (missing values always compare smaller than the smallest nonmissing value).

See Comparing Multiple B-Splines for an example of using the FILTER statement.

Filtering results has no affect on the item store contents that are displayed with the SHOW statement. However, BY-group selection with the WHERE statement can limit the amount of information that is displayed by the SHOW statements.