Filter for PLS Coefficients

The Filter for PLS Coefficients field is an optional parameter that enables you to specify a logical expression for subsetting predictors based on their estimated Partial Least Squares coefficients.

This option is useful when you want to reduce the PLS regression model to include predictors satisfying certain constraints.

The expression must be a valid SAS WHERE clause, and it must contain variable names from the plsv data set, e.g. pls1, pls2, plsnorm1, or vip. For example, to keep predictors with nonnegative weights in the second PLS component, you could use the following WHERE expression:

WHERE pls2 > 0

Note: Note: the word where has already been entered for you.

For more information about using filter fields, see The SAS WHERE Expression.