(a)
Univariate F Ratios: This will tell you if the
groups differ on one dependent variable at a time.
However, remember that the dependent variables are
usually correlated; therefore, this would result in
confounded results.
(b)
Stepdown F Ratios: Similar to above, but without
the counfounded results. All dependent variables are
prioritized from most important theoretically to
least. The most important variable is considered
first without correcting for the lower priority
variables. All subsequent variables are tested after
removing the effects of the higher priority variables
(by specifying the higher priority variables as
covariates).
(c)
Discriminant Weights: The absolute values of the
weights (the unstandardized d1
or the standardized D1) can
be used to see which variable is most influential.
However, like in Multiple Regression, these are not
stable estimates of population parameters.
(d) Structure
Correlations: For each linear composite we can
construct with MANOVA, each participant receives a
score on that dimension. To determine how much a
single dependent variable contributes to the
dimension, simply correlate all dependent variables
with the scores the individual receives on each
composite dimension. These are typically the most
stable indicator of relative contribution. The
general notion is that the higher the correlation
between the original variables and the composite
scores, the more that variable is contributing to
group differentiation.
Plot the Groups on the New
Composite Dimensions: You could plot all
individual participants on the composites to give
you a graphic depiction of how well the
dimensions discriminate amongst the groups (in
this example, the numbers 1, 2, and 3 refer to
group membership). This will help to identify,
visually, what is happening in the data.