Graphics Reference
In-Depth Information
here are two chapters on linking. Adalbert Wilhelm describes a formal model
for linked graphics and the conceptual structure underlying it. He is able to encom-
pass different types of linking and different representations. Graham Wills looks at
linking in a more applied context and stresses the importance of distinguishing be-
tween views of individual cases and aggregated views. He also highlights the variety
of selection possibilities there are in interactive graphics. Both chapterspoint out the
value of linking simple data views over linking complicated ones.
he final chapter in this section is by Simon Urbanek. He describes the graphics
that have been introduced to support tree models in statistics. he close association
betweengraphics andthemodels(andcollections ofmodelsinforests)isparticularly
interesting and has relevance for building closer links between graphics and models
in other fields.
Summary and Overview; Part III
1.2.2
he middle and largest section of the Handbook concentrates on individual area of
graphics research.
Geographical data canobviouslybenefitfromvisualization. MuchofBertin'swork
was directed at this kind of data. Juergen Symanzik and Daniel Carr write about mi-
cromaps (multiple small images of the same area displaying different parts of the
data) and their interactive extension.
Projection pursuit and the grand tour are well known but not easy to use. Despite
theavailability ofattractive freesotware,itisstilladi
culttasktoanalyse datasets in
depth with this approach. Dianne Cook, Andreas Buja, Eun-Kyung Lee and Hadley
Wickhamdescribetheissuesinvolved andoutline someoftheprogressthat hasbeen
made.
Multidimensionalscalinghasbeenaroundforalongtime.MichaelCoxandTrevor
Cox (norelation, but an MDS would doubtless place them close together) reviewthe
current state of research.
Advances in high-throughput techniques in industrial projects, academic studies
and biomedical experiments and the increasing power of computers for data collec-
tion have inevitably changed the practice of modern data analysis. Real-life datasets
become larger and larger in both sample size and numbers of variables. Francesco
Palumbo, Alain Morineau and Domenico Vistocco illustrate principles of visualiza-
tion for such situations.
Some areas of statistics benefit more directly fromvisualization than others. Den-
sityestimationishardtoimaginewithoutvisualization.MichaelMinnotte,SteveSain
andDavidScottexamineestimationmethodsinuptothreedimensions.Interestingly
there has not been much progress with density estimation in even three dimensions.
Sets of graphs can be particularly useful for revealing the structure in datasets
and complement modelling efforts. Richard Heibergerand Burt Holland describe an
approachprimarily making useof Cartesian productsand the Trellisparadigm. Wei-
Yin Loh describes the use of visualization to supportthe use of regression models, in
particular with the use of regression trees.
