SIT1610-Social Network Analysis –Unit 3

UNIT 3 CLIQUES, CLUSTERS AND COMPONENTS 9 Hrs.

Components and Subgraphs: Sub graphs - Ego Networks, Triads, Cliques, Hierarchical
Clustering, Triads, Network Density and conflict. Density: Egocentric and Sociocentric -
Digression on Absolute Density – Community structure and Density, Centrality : Local and
Global - Centralization and Graph Centres, Cliques and their intersections, Components and
Citation Circles - Positions, Sets and Clusters.

Components and Subgraphs

A subgraph is a subset of the nodes of a network, and all of the edges linking
these nodes. Any group of nodes can form a subgraph—and further down we
will describe several interesting ways to use this.
Component subgraphs (or simply components) are portions of the network that
are disconnected from each other. Before the meeting of Romeo and Juliet, the
two families were quite separate (save for the conflict ties), and thus could be
treated as components.
Many real networks (especially these collected with random sampling) have
multiple components. One could argue that this is a sampling error (which is
very possible)—but at the same time, it may just mean that the ties between
components are outside of the scope of the sampling and may in fact be
irrelevant.
Blocks and Cutpoints (Bi-components)
An alternative approach to finding the key "weak" spots in the graph is to ask: if
a node were removed, would the structure become divided into un-connected
parts? If there are such nodes, they are called "cutpoints." And, one can imagine
that such cutpoints may be particularly important actors -- who may act as
brokers among otherwise disconnected groups. The divisions into which cut-
points divide a graph are called blocks. We can find the maximal non-separable
sub-graphs (blocks) of a graph by locating the cutpoints. That is, we try to find
the nodes that connects the graph (if there are any). Another name for a block is
a "bi-component."
Factions
Imagine a society in which each person was closely tied to all others in their own
sub-population (that is, all sub-populations are cliques), and there are no
connections at all among sub-populations (that is, each sub-population is a
component). Most real populations do not look like this, but the "ideal type" of
complete connection within and complete disconnection between sub-groups is a
useful reference point for assessing the degree of "factionalization" in a
population.If we took all the members of each "faction" in this ideal-typical
society, and put their rows and columns together in an adjacency matrix (i.e.
SIT1610-Social Network Analysis –Unit 3

permuted the matrix), we would see a distinctive pattern of "1-blocks" and "0-
blocks." All connections among actors within a faction would be present, all
connections between actors in different factions would be absent.

Network Measurements
A number of measurable network characteristics were developed to gain a
greater insight into networks, with many of them having their roots in social
studies on the relationships among social actors. In this section, we will discuss
three categories of measurements that have been defined in the social network
analysis stream:
1. Network connection, which includes transitivity, multiplexity, homophily,
dyads and mutuality, balance and triads, and reciprocity
2. Network distribution, which includes the distance between nodes, degree cen-
trality, closeness centrality, betweenness centrality, eigenvector centrality and
density
3. Network segmentation, which includes cohesive subgroups, cliques, clustering
coefficient, k-cores, core/periphery, block models, and hierarchical clustering

Network Connection
Network connection (or connectivity) refers to the ability to move from one node
to another in a network. It is the ratio between route distance and geodesic
distance. Connectivity can be calculated locally (for a part of the network) and
globally (for the entire network). Let’s take a look at some of the important
metrics of network connection.

Fig.3.1 Transitivity between nodes

Transitivity
SIT1610-Social Network Analysis –Unit 3

Transitivity is a network property that refers to the extent to which a relation

between two nodes is transitive. It is a very important measure in social networks
but less important in other types of networks. In social networks, the term
transitivity reflects the friend-of-a-friend concept. It is sometimes used as a
synonym of whole-network clustering coefficient.

Homophily

Homophily is the tendency of individuals to connect with others who share the
same attitudes and beliefs. The tendency of individuals to associate with similar
others based on gender, education, race, or other socioeconomic characteristics is
very common in social communities. Coordination and cooperation are typically
more successful between people who show some similarity to each other such
that individuals in homophilic relationships are likely to hear about new ideas or
ask for help from each other.Homophily in the context of online social
networking can be understood from the similarity of users who are using the
network in terms of age, educational back- ground, region, or profession. In the
sense of corporate networks, homophily is translated as the similarity of
professional or academic qualifications.

Balance and Triads

A triad is a network structure consisting of three actors and three dyads. Given a
complete graph of three actors (triad), we can identify four different types of
rela- tionships, depending on the number of negative relationships between
nodes: (a) a friend of my friend is my friend, (b) an enemy of my enemy is my
friend, (c) a friend of my friend is my enemy, and (d) an enemy of my enemy is
my enemy.Triads with an odd number of “+” edges are bal- anced, while triads
with an even number of “-” edges are unbalanced. Imbalanced graph
configurations usually create stress for individuals located on them.

Fig. 3.2 Graph with three nodes in four states

SIT1610-Social Network Analysis –Unit 3

The above graph is a type of signed graphs which have been studied since the
1950s. They are a special case of valued graphs in which ties are allowed to have
one of two opposing values to convey the positive or negative sentiment.
Examples of signed graphs include friend/foe, trust/distrust or like/dislike,
esteem/disesteem, praise/blame, influence/negative influence, etc. They are very
common in sociology and psychology but less common in fields such as physics
and chemistry.
• In figure a, all the three actors have positive feelings, and there is no place for
conflict among them. The configuration is coherent and lacks inner tensions
between members.
• Figure b is also stable since two actors (B and C) share the same negative
feeling towards actor A, but they like each other.
• Figure c is unstable because actors A and B have a negative feeling towards
each other, while both have a positive feeling towards actor C which has to
divide its loyalty between the other two actors.
• Figure d is also unstable and will eventually break down, as it has an odd
number of negative signs.
In Fig. 3.2, b, types of balanced subgraphs are shown, whereas in Fig. 3.2 c, d,
types of unbalanced graphs are presented. An obvious way to avoid unbalances
in subgraphs is by sign shifting, which includes changing signs such that
enmities (negative signs) become friendships (positive signs) or vice versa.
Within real net- works, stable configurations appear far more often than unstable
configurations.It should be noted here that a negative sign between two nodes
does not mean the lack of tie between these two nodes. While a negative sign
between two nodes is a clear mark of an inimical relationship, the absence of a tie
between these nodes suggests the absence of interaction or communication
between them.
Reciprocity
Reciprocity is a measure of the tendency towards building mutually directed con-
nections between two actors. It refers to the number of reciprocated tie for a
specific actor in a network. For example, if u connects to v, then v connects to u
and vice versa. In real life scenarios, it is important to know whether received
help is also given or whether given help is translated as help by the receiver.For a
given node v, reciprocity is the ratio between the number of nodes which have
both incoming and outgoing connections from/to v, to the number of nodes
which only have incoming connections from v. For an entire network, reciprocity
is calculated as the fraction of edges that are reciprocated. Average reciprocity is
SIT1610-Social Network Analysis –Unit 3

calculated by averaging reciprocity values of all nodes in the network.

Network Distribution
Measurements of network distribution are related to how nodes and edges are
distributed in a network.
Distance Between Two Nodes
Distance is a network metric that allows the calculation of the number of edges
between any pair of nodes in a network. Measuring distances between nodes in
graphs is critical for many implementations like graph clustering and outlier
detection. Sometimes, the distance measure is used to see if the two nodes are
similar or not. Any commonly used shortest path calculation algorithm (e.g.,
Dijkstra) can be used to provide all shortest paths in a network with their
lengths.We can use the distance measure to calculate node eccentricity, which is
the maximum distances from a given node to all other nodes in a network. It is
also possible to calculate network diameter, which is the highest eccentricity of
its nodes and thus represents the maximum distance between nodes.In most
social networks, the shortest path is computed based on the cost of tran- sition
from one node to another such that the longer the path value, the greater the
cost.Within a community, there might be many edges between nodes, but
between communities, there are fewer edges.

Degree Centrality
In degree centrality metric, the importance of a node is determined by how many
nodes it is connected to. It is a measurement of the number of direct links to
other actors in the network. This means that the larger the number of adjacent
nodes, the more important the node since it is independent of other actors that
reach great parts of the network. It is a local measure since its value is computed
based on the number of links an actor has to the other actors directly adjacent to
it. Actors in social net- works with a high degree of centrality serve as hubs and
as major channels of information.In social networks, for example, node degree
distribution follows a power law distribution, which means that very few nodes
have an extremely large number of connections. Naturally, those high-degree
nodes have more impact in the network than other nodes and thus are considered
more important. A node i’s degree central- ity d(i) can be formulated as
SIT1610-Social Network Analysis –Unit 3

where mij = 1 if there is a link between nodes i and j and mij = 0 if there is no
such link. For directed networks, it is important to differentiate between the in-
degree centrality and the out-degree centrality.
Identifying individuals with the highest-degree centrality is essential in
network analysis because having many ties means having multiple ways to fulfill
the require- ments of satisfying needs, becoming less dependent on other
individuals, and hav- ing better access to network resources. Persons with the
highest-degree centrality are often third parties and deal makers and able to
benefit from this brokerage. For directed networks, in-degree is often used as a
proxy for popularity .The figure shows that node A and node B are at exceptional
structural positions. All communications lines must go through them. This gives
us a conclu- sion that both nodes, A and B, are powerful merely because of their
excellent posi- tions. However, such a finding is largely based on the nature of
links and the nature of embedded relationships.

Fig. 3.3 Degree centrality of nodes

Closeness Centrality

Closeness centrality can be defined as how close, to a particular actor, other

actors are. It is the sum of the geodesic distances of a node to all other nodes in
the net- work. It computes the length of paths from one actor to other actors in
the network.
That actor can be important if it is relatively close to the remaining set of
actors in the network. The mathematical representation of closeness centrality,
C(i), is given as follows:
SIT1610-Social Network Analysis –Unit 3

where dij is the geodesic distance from node i to node j (number of links in the
shortest path from node i to node j).
Closeness centrality is important to understand information dissemination in
net- works in the way that the distance between one particular node and others
has an effect on how this node can receive from or send information (e.g.,
gossip) to other nodes. In social networks, this ability is limited by what is called
“horizon of observ- ability” which states that individuals have almost no sight into
what is going on after two steps.Because closeness centrality is based on the
distance between network nodes, it can be considered the inverse of centrality
because large values refer to lower cen- trality, whereas small values refer to
high centrality. Computationally, the value of C(i) is a number between 0 and 1,
where higher numbers mean greater closeness (lower average distance) whereas
lower numbers mean insignificant closeness (higher average distance) .In the
figure, the nodes in gray are the most central regarding closeness because they
can reach the rest of nodes in the network easily and equally.

Fig3.4 Closeness centrality of nodes

They have the ability to reach all other nodes in the fastest amount of time. The
other nodes lack these privileged positions.Because closeness centrality is based
on shortest path calculations, its usefulness when applied to large networks can
be brought into question in the way that closeness produces little variation in the
results, which makes differentiating between nodes more difficult.In information
networks, closeness reveals how long it takes for a bit of information to flow from
one node to others in the network. High-scoring nodes usually have shorter paths
to the rest of nodes in the network.
SIT1610-Social Network Analysis –Unit 3

Betweenness Centrality

Betweenness centrality can be described as how important an actor is, as a link

between different networks. It represents the number of times an actor needs to
pass via a given actor to reach another actor. Nodes with high betweenness
centrality control the flow of information because they form critical bridges
between other actors or groups of actors. Betweenness centrality of node i is
calculated as follows:

where gjk is the number of shortest paths from node (j) to node k (j and k ≠ i)
and gjik is the number of shortest paths from node (j) to node k passing through
the node (i).

Eigenvector Centrality

Eigenvector centrality measurement describes the centrality of a person with

regard to the global structure of the network. It assigns relative scores to all
nodes in the network based on the concept that connections to nodes with high
scoring contribute more to the score of the node in question than connections to
nodes with low scoring.It measures the extent to which a node is connected to
well-connected nodes. It is computed by taking the principal eigenvector of the
adjacency matrix. Calculating centrality in the way that eigenvector measure
proposes differs from the way that degree measure applies to calculate centrality
which is based on simply adding up the number of links of each node.
Density

Density is defined as the degree to which network nodes are connected one to
another. It can be used as a measure of how close a network is to complete. In
the case of a complete graph (a graph in which all possible edges are present),
density is equal to one. In real life, a dense group of objects has many
connections among its entities (i.e., has a high density), while a sparse group has
few of them (i.e., has a low density).
Formally, the density D(G) of graph G is defined as the fraction of edges in G
to the number of all possible edges. Density values range between zero and one
SIT1610-Social Network Analysis –Unit 3

[0, 1].
Proportion of ties between alters comparedto number possible
ties

Fig.3.5 Ties
Total ties=1
No:of possible ties=6 (N*(N-1)/2)//N is the number of nodes
Density=1/6

Cohesive Subgroups

Cohesive groups are communities in which the nodes (members) are connected
to others in the same group more frequent than they are to those who are outside
of the group, allowing all of the members of the group to reach each other.
Within such a highly cohesive group, members tend to have strong homogenous
beliefs. Connections between community members can be formed either through
personal contacts (i.e., direct) or joint group membership (i.e., indirect). As such,
the more tightly the individuals are tied into a community, the more they are
affected by group standards.

Cliques
A clique is a graph (or subgraph) in which every node is connected to every other
node. Socially translated, a clique is a social grouping in which all individuals
know each other (i.e., there is an edge between each pair of nodes). A triangle is
an exam- ple of a clique of size three since it has three nodes and all the nodes
are connected.A maximal clique is a clique that is not a subset of any other clique
in the graph. A clique with size greater than or equal to that of every other clique
in the graph is called a maximum clique.
Relaxation of Strict Cliques
• Distance (length of paths)
SIT1610-Social Network Analysis –Unit 3

• N-clique, n-clan, n-club

• Density (number of ties)
• K-plex, ls-set, lambda set, k-core, component
N-cliques
The strict clique definition (maximal fully-connected sub-graph) may be too
strong for many purposes. It insists that every member or a sub-group have a
direct tie with each and every other member. You can probably think of cases of
"cliques" where at least some members are not so tightly or closely connected.
There are two major ways that the "clique" definition has been "relaxed" to try to
make it more helpful and general.
One alternative is to define an actor as a member of a clique if they are
connected to every other member of the group at a distance greater than one.
Usually, the path distance two is used. This corresponds to being "a friend of a
friend." This approach to defining sub-structures is called N-clique, where N
stands for the length of the path allowed to make a connection to all other
members.
N-Clans
The N-clique approach tends to find long and stringy groupings rather than the
tight and discrete ones of the maximal approach. In some cases, N-cliques can be
found that have a property that is probably undesirable for many purposes: it is
possible for members of N-cliques to be connected by actors who are not,
themselves, members of the clique. For most sociological applications, this is
quite troublesome.
To overcome this problem, some analysts have suggested restricting N-cliques
by insisting that the total span or path distance between any two members of an
N-clique also satisfy a condition. The additional restriction has the effect of
forcing all ties among members of an n-clique to occur by way of other members
of the n-clique.

Ego Net
The ego-net approach to social network analysis, which takes discrete
individual actors and their contacts as its starting point, is one of the most widely
used approaches.
 Ego network (personal network)
 Ego: focal node/respondent
 Alter: actors ego has ties with
 Ties between alters
SIT1610-Social Network Analysis –Unit 3

Fig 3.6 Ego network

Why use ego network data?
 From ego’s perspective, personal network is important for:
 Social support
 Access to resources
 Influence/normative pressure
 From a more global perspective, ego network data are useful for:
 Studying mixing patterns between groups
 Potential for diffusion
 Disease propagation
 Adoption of innovation: new product or health practice
 Lots can be had from ego network data!
 Composition of individual’s local social world
 Demographic characteristics of alters
 Shared health behaviors
 Structural features
 Size
 Density
 Nature of the ties
 Frequency, duration, closeness
 Specific exchanges

When to use Ego Network Analysis

• If your research question is about phenomena of or affecting
individual entities across different settings (networks) use the ego-
centric approach
– Individual people, organizations, nations, etc.

• If your research question is about different patterns of

interaction within defined groups (networks), use the socio-
centric approach
SIT1610-Social Network Analysis –Unit 3

– E.g., who are the key players in a group? How do ideas diffuse
through a group?

Which Theories are Ego-centric?

• Most theories under the rubric of social capital are
ego-centric
• Topological
– Structural holes / Brokerage
– Embeddedness
• Compositional
– Size
– Alter attributes

Steps to a SNA study

1. Identify the population
• Sampling, gaining access
2. Determine the data sources
• Surveys, interviews, observations, archival
3. Collect the data
• Instrument design
Step 1.Identify the Population
Sampling Criteria
– Determined by research question
• High tech entrepreneurs
• Alumni of defunct organizations
• Basketball coaches
• First time mothers returning to the workforce
• Baseball Hall of Fame inductees
• Contingent workers
• People with invisible stigmatized identities
Step 1.Identify the Population
Gaining Access
– Same concerns as other research
• It depends on the sensitivity of the questions that you are
asking
• Length of interview can be daunting
– Depends on the number of alters
SIT1610-Social Network Analysis –Unit 3

Step 2: Determine Data Sources

• Surveys
• Interviews
• Observations
• Archival data
Step 3: Collect the Data
• What data should you collect?
– What questions need to be answered?
• How to format your data collection instrument (e.g., a survey,
spreadsheet, database, etc.)?

Data Collection in an Ego-centric Study

1. Attributes about Ego

2. Name generator
• Obtain a list of alters
3. Name interpreter
• Assess ego’s relationships with generated list of alters?
4. Alter Attributes
• Collect data on the list of alters
5. Alter – Alter Relationships
• Determine whether the listed alters are connected

Attributes about Ego

• Typical variables for case based analysis
– Age
– Gender
– Education
– Profession
– SES
– Etc.
Sample Name Generators
Questions that will elicit the names of alters

– From time to time, most people discuss important personal

matters with other people. Looking back over the last six
months who are the people with whom you discussed an
important personal matter? Please just telI me their first
names or initials.
SIT1610-Social Network Analysis –Unit 3

Consider the people with whom you like to spend your free time.
– Over the last six months, who are the one or two people you
have been with the most often for informal social activities
such as going out to lunch, dinner, drinks, films, visiting
one another’s homes, and so on?

Sample Name Interpreter

• Questions that deal with ego’s relationship with [or perception
of] each alter

– How close are you with <alter>?

– How frequently do you interact with <alter>?
– How long have you known <alter>?

• All of these questions will be asked for each alter named in the
previous section

Sample Alter Attribute Questions As far as you know, what is <alter>’ s highest
 As far as you know, what is <alter>’ s highest level of education?
– Age, occupation, race, gender, nationality, salary, drug use
habits, etc
• Some approaches do not distinguish betweenname interpreters and alter
attribute
Sample Alter-Alter Relationship Questions

• Think about the relationship between <alter1> and <alter2>.

Would you say that they are strangers, just friends, or
especially close?
• Note: this question is asked for each unique alter- alter pair. E.g., if
there are 20 alters, there are 190 alter-alter relationship questions!
– Typically, we only ask one alter-alter relationshipquestion

Why Ego-Centric Analysis

• Asks different questions than whole networkanalysis.

• In fact, many of the various approaches to “Social Capital” lend

themselves particularly to the analysis of Ego-Centric or
Personal networks
SIT1610-Social Network Analysis –Unit 3

Kinds of Analyses
• In Ego-Centric Network analyses we are typically looking
to use network-derived measures as variables in more
traditional case-based analyses
– E.g., instead of just age, education, and family SES to predict
earning potential, we might also include heterogeneity of
network or brokerage statistics
Many different kinds of network measures, the simplest is degree (size
Data Analysis of Ego Networks

1. Size
– How many contacts does Ego have?
2. Composition
– What types of resources does ego have access to? (e.g., quality )
– Does ego interact with others like him/herself? (e.g., homophily)
– Are ego’s alters all alike? (e.g., homogeneity?)
3. Structure
– Does ego connect otherwise unconnected alters? (e.g.,
brokerage, density, etc)
– Does ego have ties with non-redundant alters (e.g.,
effective size,efficiency, constraint)

Size

Degree = 7

Composition: Content
• The attributes (resources) of others to whom I am connected
affect my success or opportunities
– Access to resources or information
SIT1610-Social Network Analysis –Unit 3

– Probability of exposure to/experience with

Composition: Similarity Between Ego & Alter
• Homophily
– We may posit that a relationship exists between some
phenomenon and whether or not ego and alters in a network
share an attribute
• Selection
– Teens who smoke tend to choose friends who also smoke
• Influence
– Overtime, having a network dominated by people
with particular views may lead to one taking on
those views
Composition: Homophily
• A CFO who surrounds herself with all finance people

• A Politician who surrounds himself with all members of the

same political party

Composition: Dissimilarity Between Ego & Alter

Heterophily
– We may posit that a relationship exists between some phenomenon and a
difference between ego and alters along some attribute
• Mentoring tends to be heterophilous with age

Composition: Homophily/Heterophily
Krackhardt and Stern’s E-I index

• E is number of ties to members in different groups (external), I is number of ties

to members of same group (internal).Varies between -1 (homophily) and +1
(heterophily)

Composition: Heterogeneity
• Similar to homophily, but distinct in that it looks not at similarity
to ego, but just among the alters
SIT1610-Social Network Analysis –Unit 3

• Diversity on some attribute may be provide access to different

information, opinions, opportunities, etc.

– My views about social welfare may be affected by the

diversity in SES present in my personal network
(irrespective of or in addition to my own SES)

Structural Analyses
• Burt’s work is particularly and explicitly ego- network based in
calculation

– My opportunities are affected by the connections that exist (or

are absent) between those to whomI am connected
Structural Holes

Fig 3.7 Structural holes

Burt’s Measures of Structural Holes

• Effective size
• Efficiency
• Constraint
SIT1610-Social Network Analysis –Unit 3

Fig 3.8 Structural holes

Effective Size

Fig 3.9 Effective Size

Effective Size of G = Number of G’s Alters – Sum of Redundancy of G’s

alters

= 6 – 1.33 = 4.67

Efficiency

Efficiency = (Effective Size) / (Actual Size)

SIT1610-Social Network Analysis –Unit 3

Fig 3.10 Efficiency

Actual Size = 6

Effective Size of G = 4.67

Efficiency = 4.67/6 = ~0.78

Constraint: The Basic Idea

• Constraint is a summary measure that taps the extent to which
ego's connections are to others whoare connected to one another.

• If ego's boyfriend bowls with her brother and father every

Wednesday night, she may be constrained in terms of distancing
herself from him, even if they break up.

• There's a normative bias in much of the literature that less

constraint is good
Ego-Centric Network Analysis
• When conducted across many, independent egos, presents
different problems

• Many Social Network Analysis tools ill suited to the nature

of such analyses
– Really designed for “whole network” analysis

• Ego Network analyses require either:

– joining into one large, sparse, blocked network, or
– repetition of analysis of individual networks
SIT1610-Social Network Analysis –Unit 3

• Can be tedious if there’s no facility for batching them

Local and Global Measures

Local: personal network size
– Number of alters (social support) predicting health outcomes
– Number of drug partners predicting future risky behavior
Global: degree distribution by aggregating over all cases
– Distribution of ties per person