Graphics Reference
In-Depth Information
men, the difference is subtle. You can confirm this more precisely if desired
by calculating graph statistics like average degree and centrality for both
menandwomen.InNodeXL,youcancalculatethesestatisticsforeachnode
in the workbook by running Graph Metrics and then subsequently analyze
them using column filters and standard Excel statistical formulas.
Tip
Visualization reveals complex patterns visually that you can confirm
more precisely through statistical calculation.
Another observation you may note is that the pink nodes are evenly
distributed across the graph. Women in this sample do not seem to be any
more likely to comment together. In fact, the opposite seems to be true—or
is it? Filtering may provide a clearer picture.
Find the Filters tab, and locate the Partition filters in the
Attributes
folder. Drag the Sex filter into the Queries pane below it. Check the “female”
box and click the Filter button to hide the other nodes.
Now you can see which female fans are connected to other female fans.
However, because the majority of nodes have been removed, you can no
longer rely on spatial layout to see whether or not they belong to the same
communities. To proceed with more advanced analysis, each node will need
to be explicitly associated with a community.
Community Detection
Computational community detection is a fertile area of scientific research.
Many algorithms have been written to identify communities, with various
strengthsandweaknessesandrepresentingawidevarietyofapproaches.All
of the following can be factors in choosing the most useful algorithm in any
given situation:
•
Desired count
—This is the approximate number of communities that
are informative and useful to display. In mathematics, this is often
expressed as the variable
k
.
•
Minimum criteria
—While count is important, communities should
still be held to a threshold of validity. If roughly seven is ideal for
display but the fourth, fifth, sixth, and seventh are communities only by
