Graphics Reference
In-Depth Information
Node Size
Node shape is only one dimension of graph expression, but clearly it is an
importantandhighlynuancedone.Nodesizeisanother,andcomparatively,
it is dead simple. You should generally use node size in any reasonably large
graph. Node size emphasizes what's most important in a graph when there
is a lot to take in.
The first rule of node size is hopefully a rather obvious one, which is that
size should only be mapped from magnitudes, with larger sizes indicating
greater magnitude. Often, node size is mapped from a derivate graph data
value like degree, which is the number of links attached to it, or a more
sophisticated (but computationally expensive) measure of significance (like
betweennesscentrality). Theminimum nodesizeinalargegraphistypically
related to the number of pixels required to display the node shape with
sufficient clarity. A node should also be a little bigger than its widest link.
The maximum node size is typically related to the number of nodes in the
graph.
A smaller graph can afford a greater range in node size, whereas, in a
very large graph, the range of size must be minimal. Even a size difference
of a few pixels in each direction, however, can make for a perceptible,
informative difference.
Technically, to be perceptually correct, the area and not the radius or height
and width of a node should be linearly proportional to the value it
represents. However, in graph visualization, very often this is less useful
than a rule of thumb about the overall distribution of node size.
Because the function of node size in a graph is to communicate relative
significance of nodes and significance is a fuzzy measure, it is more
important to be able to see size relative to others than it is to be able to
visually decode it precisely to a value. A reasonable rule of thumb is that
the number of first-class nodes that can be easily perceived to be most
important is 25 or less, and the number of perceptible second-class nodes
is not much more than 100. The goal is to be able to make out the key
nodes individually in the general mass context of communities. To achieve
the target distribution, you can use non-linear scales.
Search WWH ::




Custom Search