DESIGN Keyword (MANOVA: Univariate command)
You can enter one or more of the following specifications
in parentheses following the keyword DESIGN
. If DESIGN
is requested without
further specification, the default is OVERALL
.
The DECOMP
and BIAS
matrices can provide valuable information
about the confounding of the effects and the estimability of the chosen
contrasts. If two effects are confounded, the entry corresponding
to them in the BIAS
matrix will
be nonzero; if the effects are orthogonal, the entry will be zero.
This feature is particularly useful in designs with unpatterned empty
cells. For further discussion of the matrices, see Bock (1985).
OVERALL. The overall reduced-model design matrix (not the contrast matrix). This setting is the default.
ONEWAY. The one-way basis matrix (not the contrast matrix) for each factor.
DECOMP. The upper triangular QR/CHOLESKY decomposition of the design.
BIAS. Contamination coefficients displaying the bias that is present in the design.
SOLUTION. Coefficients of the linear combinations of the cell means that are used in significance testing.
REDUNDANCY. Exact linear
combinations of parameters that form a redundancy. This
keyword displays a table only if QR
(the default) is the estimation method.
COLLINEARITY. Collinearity
diagnostics for design matrices. These diagnostics include
the singular values of the normalized design matrix (which are the
same as those values of the normalized decomposition matrix), condition
indexes corresponding to each singular value, and the proportion of
variance of the corresponding parameter accounted for by each principal
component. For greatest accuracy, use the QR
method of estimation whenever you request collinearity
diagnostics.
ALL. All available options.