Graphics Reference
In-Depth Information
Insummary,bothageandeducation clearlymatterhere.Butistherereallyanysig-
nificant influence of memory on attitude towards punishment apart from this? Or,
rephrasedasahypothesis:arememoryandattitude conditionally independent,given
age and education? To focus on this question, we take another model-based view of
the data whichavoids the distracting influence of the marginal variables (ageanded-
ucation). We employ partial mosaicplots in a trellis layout (conditioning on age and
education) with residual-based shading and data-driven cut-offs(see Fig.
.
).his
visualization illustrates much more clearly that the increased approval rate among
people that recall being punished is present mainly in the elementary education col-
umn and, to a lesser degree, in the age
+ row. Using a permutation test similar to
theone discussedinSect.
.
.
,modelstatistics reveal that thehypothesis isrejected
atasignificance levelof
%,buttheshading clearlyshowsthat this association isonly
significant in two cells (the two older age groups with elementary education), where
Figure
.
.
Conditional mosaicplot for the punishment data. For each mosaicplot, rows correspond
to memory (first split) and columns to attitude (second split)
