The HPREDUCE Procedure

The HPREDUCE procedure is a high-performance procedure that performs both supervised and unsupervised variable selection on the SAS appliance. You can use the HPREDUCE procedure to perform variable selection.

The HPREDUCE procedure performs unsupervised variable selection by identifying a set of variables that jointly explain the maximum amount of data variance. Unlike principal component analysis (PCA), which reduces dimensionality by generating a set of new variables (variable extraction), the HPREDUCE procedure reduces dimensionality by selecting a subset of the original variables (variable selection). Thus, this technique preserves model interpretation.

The HPREDUCE procedure performs supervised variable selection by identifying a set of variables that jointly explain the maximum amount of variance contained in the response variables. The HPREDUCE procedure supports variable selection in both the regression setting and the classification (categorization) setting.

The HPREDUCE procedure can also be used to output the sums of squares and crossproducts (SSCP) matrix, the correlation (CORR) matrix, or the covariance (COV) matrix for exploratory data analysis and direct input to statistical procedures that accept that form. This step saves time by eliminating redundant matrix aggregations.