NE EN

VARIANCE EXPLAINED BY CANONICAL VARIABLES OF DEPENDENT VARIABLES
CAN. VAR, PCT VAR DEP CUM PCT DEP PCT VAR COV CUM PCT COV
1 T9. 14597 T9. 14597 39. 48249 39. 48249

CORRELATIONS BETWEEN COVARIATES AND CANONICAL VARJABLES

CAN. VAR.
COVARIATE 1
INTEL - 94646
CONOBV 30260
CONRUT .56188
JOB 57787
cil 62672
CI2 -69288
CI3 -79114

VARIANCE EXPLAINED BY CANONICAL, VARIABLES OF THE COVARIATES
CAN. VAR. PCT VAR DEP CUM PCT DEP PCT VAR COV CUM PCT COV
1 22.34706 22. 34T06 44. 79657 44.79657

Note that although the number of canonical variates is egual to s — min(p,g), MANOVA prints
only those variates that have a significant canonical correlation. The default significance level is
0.15 and can be changed by using the ALPHA specification, as described in Section 1.35.

1.42 Residuals ž

MANOVA will calculate and print predicted values and residuals for each response variable if
PRINT<POBS/

is specified in a MANOVA run (POBS stands for predicted observation). The output also includes
the case numbers, observed values, and standardized residuals (obtained by dividing the residuals
by the error standard deviation).

If multiple error terms are specified in an analysis of covariance model and the residuals for
each case are needed, the ERROR subphrase should be used to designate which error term's
regression coefficients are to be used in calculating the predicted values. Any error term defined in
the design can be used. Consider, for example, a 3 X 2 factorial design with repeated measures on
factor B, a SUBJECT factor nested within factor A, and a covariate X. (See Section 1.44 for a
discussion of the repeated measures design.) The following MANOVA cards may be used to obtain
residuals and significance tests for the model.

MANOVA Y BY A(1,3) SUBJECT(1.3) B(l,2) WITH X/
PRINT-POBS( ERROR (2) ) /

DESIGN<A VS l, B VS 2, A BY B VS 2,
SUBJECT W A < l, B BY SUBJECT W A - 2/

ERROR(2) within the POBS phrase indicates that the regression coefficients associated with error
term 2 are to be used to calculate the predicted values for the model (error term 2 is defined in the
DESIGN specification as the interaction between B and SUBJECT (within A)).

Various residual plots (observed versus predicted values, observed values versus standardized
residuals, predicted values versus standardized residuals, and case number versus standardized
residuals) are also available. For a discussion of the graphic features of MANOVA see Section 1.51.

1.43 SPECIAL TOPICS

1.44 Repeated Measures Designs

1.45 Introduction :

Designs in which multiple observations are made on a single experimental unit are called repeated
measures designs. For example, if a patient's blood pressure is recorded daily for five days after
administration of antihypertensive are medication, five repeated observations are obtained for the
same case. If only one variable is being measured, say systolic blood pressure, the design is termed
singly multivariate. If several variables, such as standing and recumbent systolic and diastolic blood
pressures are recorded, the design is doubly multivariate. Since multiple observations are made on
the same experimental unit, they are not independent. Special procedures must therefore be used
for analysis of repeated measures data.

MANOVA

47