Linked Activity Spaces: Embedding Social Networks in Urban Space - Computational Approaches for Urban Environments - page 316

Additionally, it is rare that an ego's alters will visit a POI that the ego does not
visit. The average proportion of total egos who have visited a specific POI (e.g.,
“flower park”) is proportional to the average percentage of their friends who have
also visited the POI. For example, consider a group of 100 egos, each of whom has
unique 50 alters (summing to a total of 5,000 friends). If an average of 50 % of
one ego's alters (25 alters) have visited a certain POI, then there is a ½ chance the
egos has visited there as well. If the average at another park is 20 % of all alters (10
alters), there is about a 1/5th chance that the egos has visited this park as well. The
chance that an ego has visited a POI is 1.036 times the number of total alters who
have visited the POI, with an r 2 correlation of 0.968. This means there are few, if
any, POIs where one frequents and his or her friends do not frequent. Conversely,
there are also few POIs where one does not frequent yet a high percentage of his or
her friends visit often.

13.4.2 Social Personas

In traditional social network analysis, a user's role in the network can reflect his or
her importance and prominence in various facets of social life, such as providing
information about new job opportunities to ones alters. For instance, a figure with
a special role in a social network (i.e., a figure with many friends, or who is a
“common friend” between poorly connected groups) can be identified through social
network metrics such as betweenness centrality, degree centrality, or brokerage
statistics (Jackson 2010 ). In one case, it has been shown that those with higher
network centrality live in more central places on the Euclidean grid of longitude
and latitude for the network (Onnela et al. 2011 ).
Confirming our second hypothesis, we find that high-degree users use the city
center more often than expected, given a random set of users. The top 1 % of high-
degree users (equating to users with 150 or more friends) concentrate at the urban
center. A Fisher-Snedecor test (F-test) of ANOVA yields a p -value of 0.006 (95 %
c.i.) signifying that the spatial variance between the high-degree users' activity
spaces is significantly lower than the spatial variance between activity spaces of the
universal population. This result illustrates high-degree agents' proclivity toward
high-density areas that are shown to be more innovative, dynamic, and energetic
environments (Bettencourt 2013 ).
We do not find significant spatial patterning with ego characteristics such as
clustering coefficient (Jackson 2010 ), which measures whether one's friends are
also friends themselves.
Using Spearman's correlation statistic, we find no significant relationship be-
tween user degree or total talking time (call duration) and the size of the user's
activity space.