Graphics Reference
In-Depth Information
errors, or populations. Even if other data attributes are not available, you
can use graph statistics such as degree and centrality. All of these additional
data attributes are not visible in the layout of the graph, but you can show
them by using a variety of visual attributes. In particular, the visual
attributes of size, color, and label are very strong visual cues that can
effectively indicate information and are configurable in most graph software
packages.
Node Size
Size is a useful visual attribute for data that represents a
magnitude
. Data
such as counts and sums in graph statistics (such as degree, that is, a count
of the number of connections) are good candidates to represent with size.
Data that is not negative or zero works well. Fields such as number of
followers, total sales, page count, market capitalization, total bytes, weight,
number of likes, number of passengers, total trade, and proportion of a
whole can all work well. These are all non-negative data examples.
If the data has zero values, consider having a minimum size instead of zero,
which is an invisible node. If the data has both positive values and negative
values (for example, profit or loss), consider setting size to the absolute
values and color to the sign (for example, positive to cyan and negative to
red).
In most software, you will have some indication of the original range of
your data values. In the e-mail data for
Figure 5-1
,
the number of e-mails
among family members ranges from 1 to 36, and the total size of e-mails
for each family member ranges from 9KB (for example, short messages) to
1,300KB (for example, messages with image attachments). Note that the
rangeisfairlysmallforthenumberofe-mails(thelargestnumberofe-mails
is 36 times the smallest), whereas it is larger for total size (the largest e-mail
is 140 times larger than the smallest). This range may impact how you
configure size.
Accurate Size
You may want people to be able to look at the graph and visually estimate
relative sizes. In this case, you want a node that is visually twice as big to
represent data that is twice as large. Note that
node size
is an
area
. This
means that a node that is twice as big in area is, in fact, only 1.4 times as
wide and high as the smaller node.
Figure 5-2
shows accurately sized nodes
based on the number of e-mails sent.
