SAVE Subcommand (DISCRIMINANT command)

SAVE allows you to save casewise information as new variables in the active dataset.

• SAVE applies only to the current analysis block. To save casewise results from more than one analysis, specify a SAVE subcommand in each analysis block.
• You can specify a variable name for CLASS and rootnames for SCORES and PROBS to obtain descriptive names for the new variables.
• If you do not specify a variable name for CLASS, the program forms variable names using the formula DSC_m, where m increments to distinguish group membership variables saved on different SAVE subcommands for different analysis blocks.
• If you do not specify a rootname for SCORES or PROBS, the program forms new variable names using the formula DSCn_m, where m increments to create unique rootnames and n increments to create unique variable names. For example, the first set of default names assigned to discriminant scores or probabilities are DSC1_1, DSC2_1, DSC3_1, and so on. The next set of default names assigned will be DSC1_2, DSC2_2, DSC3_2, and so on, regardless of whether discriminant scores or probabilities are being saved or whether they are saved by the same SAVE subcommand.
• The keywords CLASS, SCORES, and PROBS can be used in any order, but the new variables are always added to the end of the active dataset in the following order: first the predicted group, then the discriminant scores, and finally probabilities of group membership.
• Appropriate variable labels are automatically generated. The labels describe whether the variables contain predictor group membership, discriminant scores, or probabilities, and for which analysis they are generated.
• The CLASS variable will use the value labels (if any) from the grouping variable specified for the analysis.
• When SAVE is specified with any keyword, DISCRIMINANT displays a classification processing summary table and a prior probabilities for groups table.
• You cannot use the SAVE subcommand if you are replacing the active dataset with matrix materials (see Matrix Output) .

CLASS [(varname)]. Predicted group membership.

SCORES [(rootname)]. Discriminant scores. One score is saved for each discriminant function derived. If a rootname is specified, DISCRIMINANT will append a sequential number to the name to form new variable names for the discriminant scores.

PROBS [(rootname)]. For each case, the probabilities of membership in each group. As many variables are added to each case as there are groups. If a rootname is specified, DISCRIMINANT will append a sequential number to the name to form new variable names.

Example

DISCRIMINANT GROUPS=WORLD(1,3)
 /VARIABLES=FOOD TO FSALES
 /SAVE CLASS=PRDCLASS SCORES=SCORE PROBS=PRB
 /ANALYSIS=FOOD SERVICE COOK MANAGER FSALES 
 /SAVE CLASS SCORES PROBS.

• Two analyses are specified. The first uses all variables named on the VARIABLES subcommand and the second narrows down to five variables. For each analysis, a SAVE subcommand is specified.
• For each analysis, DISCRIMINANT displays a classification processing summary table and a prior probabilities for groups table.
• On the first SAVE subcommand, a variable name and two rootnames are provided. With three groups, the following variables are added to each case:

  Table 1. Saved variables based on rootnames

  Name	Variable label	Description
  PRDCLASS	Predicted group for analysis 1	Predicted group membership
  SCORE1	Function 1 for analysis 1	Discriminant score for function 1
  SCORE2	Function 2 for analysis 1	Discriminant score for function 2
  PRB1	Probability 1 for analysis 1	Probability of being in group 1
  PRB2	Probability 2 for analysis 1	Probability of being in group 2
  PRB3	Probability 3 for analysis 1	Probability of being in group 3

• Since no variable name or rootnames are provided on the second SAVE subcommand, DISCRIMINANT uses default names. Note that m serves only to distinguish variables saved as a set and does not correspond to the sequential number of an analysis. To find out what information a new variable holds, read the variable label, as shown in the following table: