Graphics Reference
In-Depth Information
hesederivative relations mightormightnotgetexposedincorrelation-based order-
ing. Semantic similarities indicate dimensions that have related meanings within the
domain; even if the values do not correlate well, users might logically group or or-
der them to help in their analysis task. Finally, some dimensions are likely to have
more importance than others for a given task, and thus ordering or assigning such
dimensions to more visually prominent features of the glyph (or, in some cases, the
features the user is likely to examine first, such as the letmost bar of a profile glyph)
will likely have a positive impact on task performance.
Glyph Layout Options
7.7
he position of glyphs can convey many attributes of data, including data values or
structure(order,hierarchy),relationships,andderivedattributes.InthissectionIwill
describe a taxonomy of glyph layout strategies, presented in detail in (Ward,
),
based on the following considerations:
Whether the placement will be data driven, e.g., based on two or more data di-
mensions, or structure driven, such as methods based on an explicit or implicit
order or other relationship between data points.
Whether overlaps between glyphs will be allowed. his can have a significant im-
pact on the size of the data set that can be displayed, the size of the glyphs used,
and the interpretability of the resulting images.
he tradeoff between optimized screen utilization, such as found in space-filling
algorithms, versus the use of white space to reinforce distances between data
points.
Whether the glyph positions can be adjusted ater initial placement to improve
visibility at the cost of distorting the computed position. Overlapping glyphs can
be di
cult to interpret, but any movement alters the accuracy of the visual de-
piction. We need to know, for the given domain, what the tradeoffs are between
accuracy and clarity.
Data-driven Placement
7.7.1
Data-driven glyph placement, as the name implies, assumes a glyph will be posi-
tioned based entirely on some or all of the data values associated with the corre-
sponding record. We differentiate two classes of such techniques based on whether
the original data values are used directly or whether positions are derived via com-
putations involving these data values. An example of the first would be the position-
ing of markers in a scatterplot using two dimensions (Fig.
.
), while an example of
the second would be to use PCA to generate the x and y coordinates of the result-
ing glyph (Fig.
.
). More complex analysis has also been used in glyph placement.
Several researchers (Globus et al.,
; Helman and Hesselink,
) have proposed
methods for placing glyphs at critical points within flow fields from fluid dynamics
