Exploring Hierarchies Using the DAGMap - Methods for Multilevel Analysis and Visualisation of Geographical Networks - page 102

Additional space is obviously required when dealing with nodes of various sizes.
Edge readability requires that layers be kept apart and that horizontal edges be
avoided. Even with medium-sized DAGs, overlapping neighboring nodes can only
be avoided at the price of a bad aspect ratio. Consequently, the identification of color
or size patterns is difficult because nodes stemming from a common ancestor are
placed in rows. Note that this is also true for the standard node-link representations
of trees because in both cases, the layout is partly devoted to the encoding of the
structure of the graph (dominance relations), leaving empty screen space between
layers.

7.2.2 Space-Filling Approaches

Space-filling approaches ( Shneiderman , 1992 ) were a solution to the visualization
of trees with node attributes at the price of illustrating less structure. Treemaps are a
visualization technique for presenting hierarchical information for two-dimensional
displays ( Shneiderman , 1992 ). TreeMaps follow a space-filling approach, mapping
the leaf nodes of a tree onto contiguous areas in a plane, with several graphical
cues reflecting attributes in the data. The area itself can be computed by taking the
attributes into account. The internal nodes of the tree are only apparent through
nesting and encode several attributes through node size or color.
This space-filling strategy radically differs from classical node-link representa-
tions for trees (see di Battista et al. , 1998 ), where much care is taken with the relative
node position, reflecting the structure of the hierarchy, as opposed to semantics in
the data (leaf nodes).
The commercial success and adoption of TreeMaps in several domains of appli-
cation corroborates the usability of TreeMaps. Improved interactivity ( Chintalapani,
Plaisant, & Shneiderman , 2004 ) and versatility ( Vliegen, van Wijk, & van der
Linden , 2006 ) often make TreeMaps an obvious choice for designing visualization
systems for hierarchical data.
Many improvements on the original TreeMap representation have been suggested
( Bruls, Huizing, & van Wijk , 2000 ; van Wijk & van de Wetering , 1999 ). In the
original TreeMap algorithm, the display is cut into alternating directions parallel to
the X-axis and Y-axis. The resulting drawings (see Fig. 7.3 ) seem to make it difficult
for users to compare two cells in the TreeMap. The TreeMap cells end up being long,
thin slices, which are difficult to read both in terms of shape and color.
In the “Squarified” TreeMap layout, the authors attempt to keep the ratio between
width and height of the rectangle close to 1 (a square). Large square areas are easier
to compare with each other (see Fig. 7.4 a). Likewise, in Voronoï TreeMaps ( Balzer
& Deussen , 2005 ), cells are represented by polygons rather than by rectangles. Two
polygons are easier to compare with each other (polygons are closer to circles and
two circles are easier to compare with each other) (see Fig. 7.4 c).
Other techniques focus on the perception of the TreeMap structure. In ( van
Wijk & van de Wetering , 1999 ), the authors draw cushions over cells to highlight