Information Technology Reference
In-Depth Information
Tests of Between-Subjects Effects
Dependent Variable: MATH_DV
Type III Sum
of Squares
df
Mean Square
F
Sig.
Source
Corrected Model
Intercept
GROUP
VERB_COV
GROUP * VERB_COV
Error
Total
Corrected Total
a
127.219
153.214
5.156
102.362
1.598
43.754
4549.000
170.972
5
1
2
1
2
30
36
35
17.446
105.052
1.768
70.185
.548
.000
.000
.188
.000
.584
25.444
153.214
2.578
102.362
.799
1.458
The group
verb_cov
interaction effect is not
statistically significant.
×
a. R Squared
=
.744 (Adjusted R Squared
=
.701)
Figure 16.12
The summary table for our custom model testing the homogeneity of
regression assumption.
The only output in which we are interested is the test of signifi-
cance of the
group
verb_cov
interaction shown in the summary table in
Figure 16.12. As can be seen in the summary table, the effect is not sta-
tistically significant. We can thus presume that the assumption of homo-
geneity of regression has not been violated and we can proceed with the
ANCOVA.
∗
16.9 PERFORMING THE ANCOVA IN SPSS
16.9.1 STRUCTURING THE COVARIANCE ANALYSIS
From the main SPSS menu select
Analyze
➜
General Linear Model
➜
Univariate.
Selecting this path will open the dialog window as shown in
Figure 16.9. We have configured it with
group
as the fixed factor,
math_dv
as the dependent variable, and
verb_cov
as the covariate.
Select the
Model
pushbutton to reach the dialog window shown
in Figure 16.10. Click
Full factorial
in the
Specify Model
area (see
Figure
16.13)
and
click
Continue
to
return
to
the
main
dialog
window.
Select the
Options
pushbutton to reach the
Options
dialog window
shown in Figure 16.14. We are going to do several things in this window.
In the top portion of the window devoted to
EstimatedMarginal Means
,
click over
group
to the panel named
Display Means for
because this will
cause SPSS to output the adjusted group means.
Then click the checkbox for
Compare main effects
and select
Bon-
ferroni
from the drop-down menu for
Confidence interval adjustment
as also shown in Figure 16.14. We are here anticipating the possibility of
obtaining a significant main effect of
group
and want to perform our
multiple comparisons tests. Because the ANCOVA assesses the differences
between the adjusted means of the groups, we cannot use the
Post Hoc
dialog window; those post hoc tests are performed on the observed
means. By checking the
Compare main effects
option we will be using a