Graphics Reference
In-Depth Information
Table
.
.
Comparison of current sotware environments
SAS
S-PLUS
R
ViSta
MANET
Mondrian
Basic functionality
Shape
Residual-based shadings
(
)
Conditional views
Interaction
Extensible design
the functionality available in these systems. he figures in this chapter have all been
produced using the R system, using the extension packages vcd (Meyer et al.,
)
andscatterplot
d (LiggesandMächler,
)(Fig.
.
only), allfreelyavailable from
the Comprehensive R Archive Network (http://CRAN.R-project.org/). he R code
used for the figures is available from http://statmath.wu-wien.ac.at/projects/vcd/.
his chapter will provide an overview of the state of the art for mosaic and associ-
ation plots, from both exploratory visualization and model-based analysis perspec-
tives. Exploratory techniques will include specialized displays for the bivariate case,
aswellaspairsplot-likedisplaysforhigh-dimensional tables. Asforthemodel-based
tools, particular emphasis will be given to methods suitable for the visualization of
conditional independence tests (including permutation tests), as well as for the visu-
alization of particular GLMs (such as log-linear models). In Sect.
.
, we start with
the simple bivariate case. Sect.
.
explains how the use of color in residual-based
shadings can support data exploration, and even promotes the methods to diagnos-
tic and model-based tools by visualizing test statistics and residuals of independence
models.InSect.
.
, we show howthe basically bivariate methods straightforwardly
extend to the multivariate case by using “flat” representations of the multiway ta-
bles. In this section, we also introduce specialized displays for conditional indepen-
dencestructures.hetechniquesareillustratedusingthree-andfour-waytables.Sec-
tion
.
concludes the chapter.
Two-Way Tables
12.2
hroughoutthissection,ourexampleswillbebasedonthehospitaldata(Wing,
)
given in Table
.
.
he table relates the length of stay (in years) of
long-term schizophrenic pa-
tients in two London mental hospitals with the frequency of visits (from relatives or
friends). he length of stay (LOS) has been categorized into
-
years,
-
years,
and more than
years. here are also three categories for the visit frequency: regu-
lar (including patients who were allowed to go home), less than monthly, and never.
Wing (
) concludes from this data that the longer the patients stayed in hospital,
the less frequently they are visited, whichcan beseen fromthe column-standardized
table (see Table
.
).
