Information Technology Reference
In-Depth Information
level of
voter and kidshome
. These represent the comparisons of the
points on our line graphs (see Figure 9.3). The first main row (top quarter
of the table) isolates those who have children in the home and who voted
in the last election. The top portion of that row compares liberals first to
moderates (the difference is 22.00,
p
=
.
000) and then to conservatives
(the difference is 26.00,
p
.001); in both cases, the mean difference is
statistically significant. The next portion of that row compares moderates
first to liberals (which has already been made) and then to conservatives;
the difference is 4.00,
p
=
=
.
875, and is not statistically significant. The last
portion compares conservatives to the other two groups; both of these
comparisons have been made above but are included by SPSS to follow
through on the format it is using.
The next quarter of the table isolates those with children in the home
who failed to vote in the last election. We can see from the comparisons
that liberals and moderates did not differ significantly but both differ from
conservatives. In like manner, we can interpret the mean differences for
participants who do not have children in the home in the lower half of the
table.
9.9.2 SIMPLE EFFECTS ANALYSES: VOTER
Figure 9.15 shows the estimated marginal means and the
Pairwise Com-
parisons
for the second simple effects analysis in which we asked SPSS to
compare those who did and did not vote in the last election at each level of
politics
and
kidshome
. There are two main horizontal rows in this output,
one for those with children in the home and another for those without
children in the home. Each is divided into three portions corresponding
to the three levels of
politics
.
The first comparison for those with children in the home is for the
liberal participants. The difference between those who voted and those
who did not vote was 1.00 and is not statistically significant (
p
=
.
791).
Moderate participants differ by 18.00, a significant difference (
p
=
.
001),
but conservatives differ by a nonsignificant 5
189). For those
with no children in the home, whether or not people voted in the last
election affected satisfaction scores for moderates and conservatives but
not for liberals.
.
00 (
p
=
.
9.9.3 MAPPING SIMPLE EFFECTS ANALYSES TO THE GRAPH
OF THE INTERACTION
Mapping the simple effects to the plot of the interaction (see Figure 9.3)
requires a little more effort than it did for what may now appear to be
the relatively simple two-way interaction of Chapter 8's example. But the
task may be accomplished one step at a time by viewing the graphs and
consulting the
Paired Comparison
tables.
We focus first on the top graph dealing with those who have children
in the home. For those who voted in the last election (the solid line in
