i
i
i
i
i
i
i
i
27.8. Examples 705
Figure 27.25. Treemap showing a filesystem of nearly one million files.
Image courtesy
Jean-Daniel Fekete
(Fekete & Plaisant, 2002),
c
2002 IEEE. (See also Plate LII.)
Treemaps use containment rather than connection to show the hierarchical
relationship between parent and child nodes in a tree (B. Johnson & Shneider-
man, 1991). That is, treemaps show child nodes nested within the outlines of
the parent node. Figure 27.25 shows a hierarchical filesystem of nearly one mil-
lion files, where file size is encoded by rectangle size and file type is encoded by
color (Fekete & Plaisant, 2002). The size of nodes at the leaves of the tree can
encode an additional data dimension, but the size of nodes in the interior does not
show the value of that dimension; it is dictated by the cumulative size of their de-
scendants. Although tasks such as understanding the topological structure of the
tree or tracing paths through it are more difficult with treemaps than with node-
link approaches, tasks that involve understanding an attribute tied to leaf nodes
are well supported. Treemaps are space-filling representations that are usually
more compact than node-link approaches.
27.8.4 Geographic
Many kinds of analysis such as epidemiology require understanding both geo-
graphic and nonspatial data. Figure 27.26 shows a tool for the visual analysis
of a cancer demographics dataset that combines many of the ideas described in
i
i
i
i
i
i
i
i
706 27. Visualization
Figure 27.26. Two matrices of linked small multiples showing cancer demographic
data (MacEachren et al., 2003),
c
2003 IEEE. (See also Plate LIII).
i
i
i
i
i
i
i
i
27.8. Examples 707
this chapter (MacEachren et al., 2003). The top matrix of linked views features
small multiples of three types of visual encodings: geographic maps showing Ap-
palachian counties at the lower left, histograms across the diagonal of the matrix,
and scatterplots on the upper right. The bottom 2 × 2 matrix, linking scatterplots
with maps, includes the color legend for both. The discrete bivariate sequential
colormap has lightness increasing sequentially for each of two complementary
hues and is effective for color-deficient people.
27.8.5 Spatial Fields
Most nongeographicspatial data is modeled as a field, where there are one or more
values associated with each point in 2D or 3D space. Scalar fields, for example
CT or MRI medical imaging scans, are usually visualized by finding isosurfaces
or using direct volume rendering. Vector fields, for example flows in water or air,
are often visualized using arrows, streamlines (McLouglin et al., 2009), and line
integral convolution (LIC) (Laramee et al., 2004). Tensor fields, such as those
describing the anisotropic diffusion of molecules through the human brain, are
particularly challenging to display (Kindlmann et al., 2000). In the next chapter,
spatial fields are discussed in detail.
Frequently Asked Questions
• What conferences and journals are good places to look for further infor-
mation about visualization?
The IEEE VisWeek conference comprises three subconferences: InfoVis (Infor-
mation Visualization), Vis (Visualization), and VAST (Visual Analytics Science
and Technology). There is also a European EuroVis conference and an Asian
PacificVis venue. Relevant journals include IEEE TVCG (Transactions on Visu-
alization and Computer Graphics) and Palgrave Information Visualization.
• What software and toolkits are available for visualization?
The most popular toolkit for spatial data is vtk, a C/C++ codebase available at
www.vtk.org. For abstract data, the Java-based prefuse (http://www.prefuse.
org) and Processing (processing.org) toolkits are becoming widely used. The
ManyEyes site from IBM Research (www.many-eyes.com) allows people to up-
load their own data, create interactive visualizations in a variety of formats, and
carry on conversations about visual data analysis.
i
i
i
i
i
i
i
i
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.