CLASSIFY Subcommand (DISCRIMINANT command)

CLASSIFY determines how cases are handled during classification.

• By default, all cases with nonmissing values for all predictors are classified, and the pooled within-groups covariance matrix is used to classify cases.
• The default keywords for CLASSIFY are NONMISSING and POOLED.

NONMISSING. Classify all cases that do not have missing values on any predictor variables. Two sets of classification results are produced, one for selected cases (those specified on the SELECT subcommand) and one for unselected cases. This is the default.

UNSELECTED. Classify only unselected cases. The classification phase is suppressed for cases selected via the SELECT subcommand. If all cases are selected (when the SELECT subcommand is omitted), the classification phase is suppressed for all cases and no classification results are produced.

UNCLASSIFIED. Classify only unclassified cases. The classification phase is suppressed for cases that fall within the range specified on the GROUPS subcommand.

POOLED. Use the pooled within-groups covariance matrix to classify cases. This is the default.

SEPARATE. Use separate-groups covariance matrices of the discriminant functions for classification. DISCRIMINANT displays the group covariances of canonical discriminant functions and Box’s test of equality of covariance matrices of canonical discriminant functions. Since classification is based on the discriminant functions and not the original variables, this option is not necessarily equivalent to quadratic discrimination.

MEANSUB. Substitute means for missing predictor values during classification. During classification, means are substituted for missing values and cases with missing values are classified. Cases with missing values are not used during analysis.