Graphics Reference
In-Depth Information
tailed discussion on log-linear models and mosaicplots can be found in heus and
Lauer (
).
Trellis Displays
6.3
Trellis displays (called Lattice Graphics within the R package) also use conditioning
toplot high-dimensional data. But whereas mosaicplots use a recursive layout, trellis
displays use a gridlike structure to plot the data conditioned on certain subgroups.
Deinition
6.3.1
Trellis displays were introduced by Becker et al. (
) as a means to visualize mul-
tivariate data (see also heus,
). Trellis displays use a latticelike arrangement to
place plots onto so-called panels. Each plot in a trellis display is conditioned upon
at least one other variable. To make plots comparable across rows and columns, the
same scales are used in all the panel plots.
he simplest example of a trellis display is probably a boxplot y by x.Figure
.
shows a boxplot of the gas mileage of cars conditioned on the type of car. Results can
easily be compared between car types since the scale does not change when visually
traversing the different categories. Even a further binary variable can be introduced
when highlighting is used, which would be the most effective way to add a third
(binary) variable to the plot.
In principle, a single trellis display can hold up to seven variables at a time. Nat-
urally five out of the seven variables need to be categorical, and two can be contin-
uous. At the core of a trellis display we find the panel plot.heuptotwovariables
plottedinthepanelplotarecalled axis variables.(hecurrentLattice Graphicsimple-
mentation in R does actually offer higher-dimensional plots like parallel coordinate
Figure
.
.
A boxplot y by x is a simple form of a trellis display
