Introduction to graph theory

Graphs
Size and order
Degree and degree distribution
Subgraphs
Paths, components
Geodesics
Some special graphs
Centrality and centralisation

Directed graphs
Dyad and triad census
Paths, semipaths, geodesics, strong and weak components
Centrality for directed graphs
Some special directed graphs

©Department of Psychology, University of Melbourne

Definition of a graph

A graph G comprises a set V of vertices and a set E of edges

Each edge in E is a pair (a,b) of vertices in V

If (a,b) is an edge in E, we connect a and b in the graph drawing of G

Example: V={1,2,3,4,5,6,7}
E={(1,2),(1,3),(2,4).
1 (4,5),(3,5),(4,5),
2 3 (5,6),(6,7)}

4 5

6 7
©Department of Psychology, University of Melbourne

1
 Size and order

The size of G is the number n of vertices in V

The order of G is the number L of edges in E

Minimum possible order is 0 (empty graph)

Maximum possible order is n(n-1)/2 (complete graph)

Size = 7, Order = 8

©Department of Psychology, University of Melbourne

Adjacency matrix for a graph

The adjacency matrix x =[xab] for G is a matrix with n rows and n colums and
entries given by:

xab = 1 if (a,b) is an edge in G

0 otherwise

Example: graph adjaceny matrix

1 0110000
1001000
2 3 1000100
0100100 symmetric
4 5 0011011
0000101
6 7 0000110

©Department of Psychology, University of Melbourne

2
 Density

The density of G is the ratio of edges in G to the maximum possible

number of edges

2L
Density = --------
n(n-1)

Density = 2×8/(7×6) = 8/21

©Department of Psychology, University of Melbourne

Degrees and degree sequence

The degree da of vertex a is the number of vertices to which a is linked by an
edge
The minimum possible degree is 0
The maximum possible degree is n-1

The degree sequence for a graph is the vector (d1, d2,…, dn)

1
2 3

4 5

6 7
Degree sequence = (2,2,2,2,4,2,2)
©Department of Psychology, University of Melbourne

3
 Degree distribution

The degree distribution for the graph is (k0, k1,…, kn-1),

where kj = the number of nodes with degree j

frequency

2 4 degree

©Department of Psychology, University of Melbourne

Subgraphs

A subgraph of G=G(V,E) is a subset W of the vertex set V together with

all of the edges that connect pairs of vertices in W

Eg if W={4,5,6,7}, the subgraph of

1
2 3

4 5 is 4 5

6 7 6 7

©Department of Psychology, University of Melbourne

4
 Subgraph counts: the dyad census

The graph G has n(n-1)/2 subgraphs of size 2

Each subgraph of size 2 comprises a pair of vertices, and the edge

between them is either present or absent:

subgraph count

D0 = n(n-1)/2 – L

D1 = L

Dyad census = (D0,D1) count of the no. of each type of dyad subgraph

©Department of Psychology, University of Melbourne

Subgraph counts: the triad census

The graph G has n(n-1)(n-2)/6 subgraphs of size 3

Each subgraph of size 3 comprises a triple of vertices, and the possible forms are:

subgraph count

T0 = (1/6)∑i,j,k (1-xij)(1-xik )(1- xjk)

T1 = (1/6)∑i,j,k (1-xij)(1-xik )xjk

T2 = (1/6) ∑ i,j,k(1-xij)xikxjk

T3 = (1/6) ∑ i,j,k xijxikxjk

Triad census = (T0,T1 ,T2,T3)

©Department of Psychology, University of Melbourne

5
 Paths
A path from vertex a to vertex b is an ordered sequence
a=v0, v1, …, vm=b
of distinct vertices in which each adjacent pair (vj-1,vj) is linked by an edge. The
length of the path is m

There is:
1

2 3 a path of length 1 from 1 to 2

a path of length 2 from 1 to 4
4 5 a path of length 3 from 1 to 4
a path of length 3 from 1 to 6
6 7

©Department of Psychology, University of Melbourne

Reachability and connectedness

If there is a path from vertex a to vertex b, a is reachable from b

If each vertex in G is reachable from each other vertex, then G is

connected

A component of G is a maximal connected subgraph (ie a connected

subgraph with vertex set W for which no larger set Z containing W is
connected)

A graph with 3 components

©Department of Psychology, University of Melbourne

6
 Geodesics

A geodesic from a to b is a path of minimum length

The geodesic distance dab between a and b is the length of the geodesic
If there is no path from a to b, the geodesic distance is infinite

For the graph

The geodesic distances are:

dAB = 1, dAC = 1, dAD = 1, dBC = 1, dBD = 2, dCD = 2

©Department of Psychology, University of Melbourne

Cycles

A cycle is an ordered sequence

a=v0, v1, …, vm=a
of vertices in which each adjacent pair (vj-1,vj) of vertices is linked by an
edge, and v0, v1, …, vm-1 are distinct . The length of the cycle is m

Cycles of length
3 4 5

©Department of Psychology, University of Melbourne

7
 Some special graphs:
trivial, empty and complete graphs

The empty graph on 5 vertices (Z5)

The complete graph on 5 vertices (K5)

©Department of Psychology, University of Melbourne

Star and cyclic graphs

A star graph on 6 vertices

A cyclic graph on 5 vertices (C5)

©Department of Psychology, University of Melbourne

8
 Trees and forests

A tree (a connected acyclic graph)

A forest (a graph with tree components)

©Department of Psychology, University of Melbourne

Bipartite graphs

A bipartite graph (vertex set can be partitioned into 2 subsets, and there
are no edges linking vertices in the same set)

A complete bipartite graph (all possible edges are present)

K1,5 K3,2

©Department of Psychology, University of Melbourne

9
 Cutpoints

A vertex is a cutpoint if its removal increases the number of components

in the graph

the vertex marked by the

red arrow is a cutpoint

the vertex marked by

the blue arrow is not

©Department of Psychology, University of Melbourne

Bridges

An edge is a bridge if its removal increases the number of components in

the graph

the edge marked by the

red arrow is a bridge

This graph has no bridges

©Department of Psychology, University of Melbourne

10
 Connectivity

The connectivity κ(G) of a connected graph G is the minimum number of

vertices that need to be removed to disconnect the graph (or make it
empty)

A graph with more than one component has connectivity 0

Graph

Connectivity 0 1 2 4

A graph with connectivity k is termed k-connected

©Department of Psychology, University of Melbourne

Edge-connectivity

The edge-connectivity λ(G) of a connected graph G is the minimum

number of edges that need to be removed to disconnect the graph

A graph with more than one component has edge-connectivity 0

Graph

Edge-
Connectivity 1 2 2 4

Connectivity 1 2 1 4
©Department of Psychology, University of Melbourne

11
 Independent and edge-independent paths

Two paths from a to b are independent if they have no nodes in common

apart from a and b e.g. paths 1-2-4-5 and 1-3-5
1

2 3

4 5

6 7

Two paths from a to b are edge-independent if they have no edges in

common e.g. paths 1-2-4-5-6 and 1-3-5-7-6

©Department of Psychology, University of Melbourne

Several theorems about connectivity

Whitney’s theorem
For any graph G, κ(G) ≤ λ(G) ≤ δ(G), where δ(G) is the minimum degree
of any vertex in G

Menger’s theorem
A graph G is k-connected if and only if any pair of vertices in G are linked
by at least k independent paths

Menger’s theorem
A graph G is k-edge-connected if and only if any pair of vertices in G are
linked by at least k edge-independent paths

For application, see Harary & White (2001)

©Department of Psychology, University of Melbourne

12
 Degree Centrality

Freeman (1979) described three measures of vertex centrality:

Degree centrality (communication potential)

Degree centrality of node a: CD(a) = da degree of node a

Normalised degree centrality of node a: da/(n-1) x

Example
Node x: degree centrality = 4
normalised degree centrality = 4/6 = 0.67
©Department of Psychology, University of Melbourne

A political network (Doreian, 1988)

©Department of Psychology, University of Melbourne

13
 Degree centrality
in Doreian’s (1988) political network
Degree NrmDegree Share
------------ ------------ ------------
4 Council 1 6.000 46.154 0.107
12 Fr Pres 6.000 46.154 0.107
3 Sheriff 5.000 38.462 0.089
8 President 5.000 38.462 0.089
6 Council 3 5.000 38.462 0.089
13 City Mayor 5.000 38.462 0.089
2 Auditor 4.000 30.769 0.071
1 Executive 4.000 30.769 0.071
9 Council 5 4.000 30.769 0.071
10 Council 6 4.000 30.769 0.071
7 Council 4 3.000 23.077 0.054
14 Prosecutor 2.000 15.385 0.036
5 Council 2 2.000 15.385 0.036
11 Fr Council 1.000 7.692 0.018

©Department of Psychology, University of Melbourne

Closeness Centrality

Closeness centrality (potential for independent communication)

Closeness centrality of node a: CD(a) = 1/∑bdab inverse sum of

distances to other
nodes b

Normalised closeness centrality of node a: (n-1)/∑bdab x

Example
Node x: closeness centrality = 1/[1+1+1+1+2+2]=1/8 = 0.125
normalised closeness centrality = 6/8=0.75
©Department of Psychology, University of Melbourne

14
 Closeness centrality in
Doreian’s political network
Farness nCloseness
------------ ------------
12 Fr Pres 20.000 65.000
4 Council 1 22.000 59.091
6 Council 3 23.000 56.522
13 City Mayor 23.000 56.522
1 Executive 24.000 54.167
8 President 25.000 52.000
3 Sheriff 26.000 50.000
2 Auditor 27.000 48.148
7 Council 4 28.000 46.429
9 Council 5 31.000 41.935
10 Council 6 31.000 41.935
11 Fr Council 32.000 40.625
5 Council 2 32.000 40.625
14 Prosecutor 32.000 40.625

©Department of Psychology, University of Melbourne

Betweeness centrality
Betweeness centrality (Potential for control of communication)

Betweenness centrality of node a: CD(a) = ∑b<c[gbc(a)/gbc]

Where gbc is the number of geodesics between b and c, and gbc(a) is the number of
geodesics between b and c that contain a

sum over all pairs (b,c) of the proportion of geodesics linking

the pair that contain node a

Normalised betweeness centrality of node a: x

2∑b<c[gbc(a)/gbc]/[n2 –3n +2]

Example
Node x: betweeness centrality = 14
normalised betweeness centrality = 14/[49-21+2]=7/15

©Department of Psychology, University of Melbourne

15
 Betweeness centrality
in Doreian’s political network
Betweenness nBetweenness
------------ ------------
12 Fr Pres 33.198 42.561
13 City Mayor 14.490 18.578
4 Council 1 13.843 17.747
6 Council 3 13.024 16.697
1 Executive 9.452 12.118
3 Sheriff 4.500 5.769
8 President 4.021 5.156
2 Auditor 3.571 4.579
9 Council 5 0.450 0.577
10 Council 6 0.450 0.577
11 Fr Council 0.000 0.000
5 Council 2 0.000 0.000
7 Council 4 0.000 0.000
14 Prosecutor 0.000 0.000

©Department of Psychology, University of Melbourne

Maximum centrality

Centrality is at a maximum on all measures for the central node in a star

configuration:

Normalised measures
Degree centrality: 1
Closeness centrality: 1
Betweeness centrality: 1

©Department of Psychology, University of Melbourne

16
 Centralisation

Graph-level measure of centralisation:

Degree to which the centrality of the most central vertex exceeds the
centrality of all other vertices, compared to the maximum possible
discrepancy

Index has the form:

Sum over nodes a of (max centrality in G – centrality of node a)

------------------------------------------------------------------------------
Max value of the sum in all graphs on the same number of vertices

Can be used with any centrality measure

©Department of Psychology, University of Melbourne

Eigenvector centrality (Bonacich, 1972)

The eigenvector centrality of vertex a is the sum of its connections to

other nodes, weighted by their centrality

It is hence given by the solution for ca of the equation

ca = (1/λ) ∑b xabcb where λ is a constant

[Mathematical note: this is equivalent to:

λc = xc
where x is the adjacency matrix and c is the vector of centrality measures;
hence c is an eigenvector of x, usually taken to be the one associated
with the largest eigenvalue λ]

Use with connected graphs only

©Department of Psychology, University of Melbourne

17
 Eigenvector centrality for Doreian network
Eigenvec nEigenvec
--------- ---------
1 Executive 0.249 35.207
2 Auditor 0.287 40.585
3 Sheriff 0.257 36.368
4 Council 1 0.328 46.358
5 Council 2 0.134 18.887
6 Council 3 0.270 38.154
7 Council 4 0.186 26.357
8 President 0.350 49.435
9 Council 5 0.282 39.854
10 Council 6 0.282 39.854
11 Fr Council 0.083 11.723
12 Fr Pres 0.371 52.454
13 City Mayor 0.343 48.453
14 Prosecutor 0.118 16.655

©Department of Psychology, University of Melbourne

Directed graph

A directed graph G comprises a set V of vertices and a set E of arcs

Each arc in E is an ordered pair (a,b) of vertices in V

If (a,b) is an arc in E, we draw an arc from a to b in the
graph drawing of G

Example: V={1,2,3,4}
E={(1,2),(2,1),(2,4),(1,3),(4,2)}
1 2

3 4

©Department of Psychology, University of Melbourne

18
 Adjacency matrix for a directed graph

The adjacency matrix x =[xab] for G is a matrix with n rows and n colums
and entries given by:

xab = 1 if (a,b) is an arc in G

0 otherwise

Example
adjacency matrix
1 2 0110
1001 not
3 4 0000 necessarily
0100 symmetric

©Department of Psychology, University of Melbourne

Density for a directed graph

The density of G is the ratio of arcs in G to the maximum possible number of arcs

L
Density = --------
n(n-1)

Example
1 2
density = 5/12

3 4

©Department of Psychology, University of Melbourne

19
 Indegrees and outdegrees

The indegree ia of vertex a is the number of vertices to which a is linked

by an arc
The outdegree oa of vertex a is the number of vertices linked to a by an
arc

The minimum possible indegree (or outdegree) is 0

The maximum possible indegree(or outdegree) is n-1

©Department of Psychology, University of Melbourne

Directed subgraphs

A subgraph of G=G(V,E) is a subset W of the vertex set V together with

all of the arcs that connect pairs of vertices in W

Eg if W={1,2,3}, the subgraph of

1 2 1 2
is
3 4 3

©Department of Psychology, University of Melbourne

20
 Directed subgraph counts: the dyad census

The directed graph G has n(n-1)/2 subgraphs of size 2

Each subgraph of size 2 comprises a pair of vertices, and there are either
0, 1 or 2 arcs linking them:

subgraph count

N = number of null dyads

A = number of asymmetric arcs
M = number of mutual arcs

Dyad census = (M,A,N) count of the no. of each type of dyadic subgraph

©Department of Psychology, University of Melbourne

Directed subgraph counts: the triad census

The graph G has n(n-1)(n-2)/6 subgraphs of size 3

Each subgraph of size 3 comprises a triple of vertices, and there are 16 possible
forms:

subgraphs

Triad census: counts of each of the 16 forms across all subgraphs of G

©Department of Psychology, University of Melbourne

21
 Paths and semipaths

A path from vertex a to vertex b is an ordered sequence

a=v0, v1, …, vm=b
of distinct vertices in which each adjacent pair (vj-1,vj) is linked by an arc. The
length of the path is m

A semipath from vertex a to vertex b is an ordered sequence

a=v0, v1, …, vm=b
of distinct vertices in which either (vj-1,vj) and/or (vj,vj-1) is linked by an arc. The
length of the semipath is m

e.g. 1 2 there is a path from 2 to 3 of length 2

there is no path from 3 to 4
3 4 there s a semipath from 3 to 4 of length 3

©Department of Psychology, University of Melbourne

Strong and weak connectedness;

strong and weak components
If there is a path from vertex a to vertex b, b is reachable from a

If each vertex in G is reachable from each other vertex, then G is strongly

connected

If there is a semipath from each vertex in G to each other vertex, then G is

weakly connected

A strong (weak) component of G is a maximal strongly (weakly)

connected subgraph (ie a strongly (weakly) connected subgraph with
vertex set W for which no larger set Z containing W is strongly
(weakly) connected)

©Department of Psychology, University of Melbourne

22
 Geodesics
A geodesic from a to b is a path of minimum length
The geodesic distance dab between a and b is the length of the geodesic
If there is no path from a to b, the geodesic distance is infinite

For the directed graph

1 2

3 4

The geodesic distances are:

d12 = 1, d13 = 1, d14 = 2, d21 = 2 d23 = 2, d24 = 1,
d31 = infinite, d32 = infinite, d34 = infinite, d41 = 2, d42 = 1, d43 = 3

©Department of Psychology, University of Melbourne

Centrality in directed graphs

As for graphs, but note that:

Indegree and outdegree centrality replace degree centrality

Eigenvector centrality is only computed by UCINET for graphs

©Department of Psychology, University of Melbourne

23
 Some special directed graphs

Empty and complete directed graphs

Cycle

Acyclic directed graph: a directed graph with no cycles

Transitive directed graph: every two path is

accompanied by a direct path

©Department of Psychology, University of Melbourne

24