Information Technology Reference
In-Depth Information
Linear Models
The GLM Procedure
satscore
Level of
group
Figure 6.17
Descriptive statistics generated by
the
Linear Models
procedure.
N
Mean
Std Dev
0
2
4
6
8
412.857143
474.285714
552.857143
614.285714
622.857143
26.9037084
47.9086432
28.7020822
38.2348632
36.3841933
7
7
7
7
7
6.9.3 HOMOGENEITY TEST
Figure 6.18 displays the result for Levene's test of homogeneity of variance.
The statistic (symbolized as
W
if you are reporting it) has a value of
1.459 and, with 4 and 30
df
, has a probability of occurrence,
p
239,
if the null hypothesis is true. This mirrors the numerical outcome we
obtained from SPSS. Against an alpha level of .05 this probability is not
statistically significant. We therefore do not reject the null hypothesis that
the variances of the five conditions are comparable; in short, it appears
that the assumption of homogeneity of variance has not been violated.
=
.
6.9.4 SUMMARY TABLE
The summary table for the omnibus analysis is presented in the top portion
of Figure 6.19. Note that SAS provides the
CorrectedTotal
sum of squares,
informing us that we are viewing the results of the reduced model. The
sum of squares associated with
Model
(see the top table in Figure 6.19) is
identical to the sum of squares associated with our
group
effect (see the
bottom table in Figure 6.19) because the only variable in the prediction
model is the group effect, that is, the
group
effect is the model. Whichever
you examine, we see a statistically significant effect numerically matching
the output from SPSS. In later chapters where we have between-subjects
designs with more than a single independent variable, the
Model
effect
will be the cumulation of all effects in the analysis, and our focus will be
on the individual effects comprising the model.
Linear Models
The GLM Procedure
Figure 6.18
The result of
the Levene test for homo-
geneity of variance.
Levene's Test for Homogeneity of satscore Variance
ANOVA of Absolute Deviations from Group Means
Sum of
Squares
Mean
Square
Pr
>
F
Source
DF
F Value
group
4
30
1929.8
9919.5
482.4
330.7
1.46
0.2393
Error
