DISCRIMINANT can read correlation matrices written by a previous DISCRIMINANT command or by other procedures.
Matrix materials read by DISCRIMINANT must contain records with ROWTYPE_ values MEAN, N or COUNT (or both), STDDEV, and CORR.
If the data do not include records with ROWTYPE_ value COUNT (unweighted number of cases), DISCRIMINANT uses information from records with ROWTYPE_ value N (weighted number of
cases). Conversely, if the data do not have N values, DISCRIMINANT uses the COUNT values. These
records can appear in any order in the matrix input file with the
following exceptions: the order of split-file groups cannot be violated
and all CORR vectors must appear
consecutively within each split-file group.
If you want to use a covariance-type matrix as input
to DISCRIMINANT, you must first
use the MCONVERT command to change
the covariance matrix to a correlation matrix.
DISCRIMINANT can use a matrix from a previous dataset to classify data in the
active dataset. The program checks to make sure that the grouping
variable (specified on GROUPS) and the predictor variables (specified on VARIABLES) are the same in the active dataset as in the
matrix file. If they are not, the program displays an error message
and the classification will not be executed.
MATRIX=IN cannot be used unless a active dataset
has already been defined. To read an existing matrix data file at
the beginning of a session, first use GET to retrieve the matrix file and then specify IN(*) on MATRIX.