Graphics Reference
In-Depth Information
Figure 5-2:
E-mails between family members are shown here with nodes
accurately sized based on number of e-mails from 1 to 36. There are a few
people who send a large number of e-mails, and the left cluster clearly has
larger nodes (more e-mails) than the right cluster.
Warning
If you want size to accurately represent data, inspect the resulting
visualization to validate the sizes by comparing different nodes. Some
software scales sizes linearly to the data (incorrect for accuracy), and
some software scales sizes to the square root of the data (correct for
accuracy).
You may want to add a column to the node data that is a square root of
the intended size data, in case your software doesn't automatically use the
squarerootofthesize.Ifyouintendtohaveaccuratesizes,itcanbeusefulto
have a legend or otherwise indicate the sizes of some of the different nodes,
including the largest and smallest nodes.
Relative Size
Relative sizes
can be useful, too, particularly where the dynamic range is
more than a couple orders of magnitude. The problem with extremely large
ranges and accurate sizes is that some nodes are teeny dots (for example, a
single pixel), and other nodes are so large that they completely overwhelm
(or obscure) other nodes, as shown in
Figure 5-3
. In this case, relative size
can be effective. With relative sizes, nodes that represent larger data are
visually bigger, and nodes that represent smaller data are visually smaller.
