The models in analysis of variance and covariance are not of full rank.
Therefore, if estimates of the effects are wanted, constraints on the
estimates must be given. Two kinds of constraints may be given by the
parameter CONSTRAINTS:
1 The estimates of a main effect or an interaction is constrained
to zero.
2 The weighted sum of estimates of a main effect or an interaction
is constrained to zero, using sample sizes as weights.
If the sample sizes are unequal then estimates can be obtained only
for main effects and two-terms interactions.
The predicted values of the dependent variable and/or residuals may
be saved in the data file. The name of the predicted variable is
given by the parameter PREDICTED and the name of the residual variable
by RESIDUAL, e.g.
PREDICTED=ESTKNOWL RESIDUAL=RESKNOWL
Note that the residuals or predicted values may be saved only in a data
file not in the edit field.
Further information:
1 = More on definitions for grouping variables
2 = More on analysis of variance for multiple factors
C = More on analysis of covariance
A = More on ANOVA operations
Help System (web edition)