1. Important
NETWORK METRICS Network
Measures:
Chapter 5
Presentation based on Hansen, D., Shneiderman, B., & Smith, M. A. (2011).
Analyzing Social Media Networks with NodeXl: Insights from a Connected World.
New York, NY: Morgan Kaufmann
Please provide acknowledgement for use as follows:
Kwon, H. (2013). “Social Network Analysis :Basics.” Lecture Presentation.
Arizona State University
2. NETWORK METRICS
Quantitative results by analyzing relative
structure of the whole networks and
individuals’ (vertices) positions within a
network
Two level of metrics
Overall graph metrics (network as a whole)
Vertex-specific metrics (individual within a
network)
3. 1. OVERALL GRAPH METRICS
1. Density: Measures “How highly connected vertices
are”
Density = # of edges/ # of all possible edges
*** # of all possible edges =n(n -1)/2 ***
Density?
Density?
4. 1. OVERALL GRAPH METRICS
2. Component:
A cluster of vertices
that are connected to
each other but separate
from other vertices in
the graph
3. Isolate = a single
vertex component
5. 1. OVERALL GRAPH METRICS
2. Component:
A cluster of vertices
that are connected to
each other but separate
from other vertices in
the graph
3. Isolate = a single
vertex component
6. 2. VERTEX-SPECIFIC METRICS
1 . Centrality
Degree: a count of the number of unique edges that are
connected to a given vertex
Betweenness: a measure of how often a given vertex lies on
the shortest path (geodesic distance) between two other
vertices. Higher betweenness centrality means that a
vertex is positioned as a bridge (or gatekeeper) between
many pairs of other vertices.
Closeness: the average distance between a vertex and
every other vertex in the network. Higher closeness
centrality means that a vertex has the shortest distance to
all others.
Eigenvector Centrality: a measure of the value of
connections that a given vertex has. If a vertex has
connections to others with high degree centralities, the
vertex shows high eigenvector centrality.
7. 2. VERTEX-SPECIFIC METRICS
2. Clustering Coef ficient: A measure of how a vertex’s friends
are connected to one another. If my friends have connections to
one another, I have a high clustering coef ficient. If they are not
connected, I have a low clustering coef ficient.
Measures of Degree and Eigenvector Centralities dif fer
between un-weighted (whether there is a edge or not) and
weighted (how valued the edge is) network.
Notas del editor
How many components? Is there an isolate? Which is the biggest component?