Graphics Reference
In-Depth Information
Some of the plots are optimized for presentation graphics (e.g., trellis displays),
others, in contrast only make sense in a highly interactive and exploratory setting
(e.g., grand tour).
he high-dimensional nature of the data problems for the discussed plots calls
for interactive controls - be they rearrangements of levels and/or variables or differ-
ent scalings of the variables or the classical linked highlighting - which put different
visualization methods together into one framework, thus further increasing the di-
mensionality.
Figure . illustrates the situation. When analyzing high-dimensional data, one
needsmorethanjustonevisualization technique.Dependingonthescaleofthevari-
ables (discrete or continuous) and the number of variables that should be visual-
ized simultaneously, one or another technique is more powerful. All techniques -
except for trellis displays - have in common that they only rise to their full power
when interactive controls are provided. Selection and linking between the plots can
bring the different methods together, which then gives even more insight. he re-
sults found in an exploration of the data may then be presented using static graph-
ics. At this point, trellis displays are most useful to communicate the results in an
easy way.
Finally, implementations of these visualization tools are needed in sotware. Right
now, most of them are isolated features in a single sotware package. Trellis displays
can only be found in the R package, or the corresponding commercial equivalent
Figure . . Diagram illustrating the importance of interactivity and linking of the high-dimensional
visualization tools in statistical graphics
Search WWH ::




Custom Search