1 of 78

Network Metrics:

Deep Dive

2 of 78

Network Metrics:

Deep Dive

3 of 78

How to find, create and think about network metrics

4 of 78

Network Metrics

Why?

Measures are tools for thinking.

5 of 78

Network Metrics

How?

Read the literature

Debate the literature

6 of 78

Network Metrics

How?

7 of 78

Network Metrics

Lesson for why to read:

8 of 78

Network Metrics

How?

9 of 78

Network Metrics

How?

10 of 78

Network Metrics

How?

11 of 78

12 of 78

Network Metrics

Where to start?

Other options

  • clustering coefficient, degree centralization degree distribution, density, diameter, efficiency, global_efficiency(G), mean degree, modularity, number of communities , number of components, transitivity, structural balance

13 of 78

Network Metrics

Where to start?

14 of 78

Network Metrics

Where to start?

15 of 78

Network Metrics

Where to start?

16 of 78

17 of 78

Network Metrics

Solution to theoretical problem

18 of 78

Analytically, most of these definitions & operationalizations of cohesion do not distinguish the social fact of cohesion from the psychological or behavior outcomes resulting from cohesion.

Def. 1:

“A collectivity is cohesive to the extent that the social relations of its members hold it together.”

What network pattern embodies all the elements of this intuitive definition?

Example: Cohesion & Clustering

Structural Cohesion

19 of 78

This definition contains 5 essential elements:

  1. Focuses on what holds the group together
  2. Expressed as a group level property
  3. The conception is continuous
  4. Rests on observable social relations
  5. Applies to groups of any size

Example: Cohesion & Clustering

Structural Cohesion

20 of 78

1) Actors must be connected: a collection of isolates is not cohesive.

Not cohesive

Minimally cohesive:

a single path connects everyone

Example: Cohesion & Clustering

Structural Cohesion

21 of 78

1) Reachability is an essential element of relational cohesion. As more paths re-link actors in the group, the ability to ‘hold together’ increases.

Cohesion increases as # of paths connecting people increases

The important feature is not the density of relations, but the pattern.

Example: Cohesion & Clustering

Structural Cohesion

22 of 78

Consider the minimally cohesive group:

D = . 25

D = . 25

Moving a line keeps density constant, but changes reachability.

Example: Cohesion & Clustering

Structural Cohesion

23 of 78

What if density increases, but through a single person?

D = . 25

D = . 39

Removal of 1 person destroys the group.

Example: Cohesion & Clustering

Structural Cohesion

24 of 78

Cohesion increases as the number of independent paths in the network increases. Ties through a single person are minimally cohesive.

D = . 39

Minimal cohesion

D = . 39

More cohesive

Example: Cohesion & Clustering

Structural Cohesion

25 of 78

Substantive differences between networks connected through a single actor and those connected through many.

Minimally Cohesive Strongly Cohesive

Power is centralized Power is decentralized

Information is concentrated Information is distributed

Expect actor inequality Actor equality

Vulnerable to unilateral action Robust to unilateral action

Segmented structure Even structure

Def 2.

“A group is structurally cohesive to the extent that multiple independent relational paths among all pairs of members hold it together.”

Example: Cohesion & Clustering

Structural Cohesion

26 of 78

Node Connectivity

As size of cut-set

0

1

2

3

Structural Cohesion:

A network’s structural cohesion is equal to the minimum number of actors who, if removed from the network, would disconnect it.

Example: Cohesion & Clustering

Structural Cohesion

27 of 78

0

1

2

3

Node Connectivity

As number of node-independent paths

Structural Cohesion:

A network’s structural cohesion is equal to the minimum number node independent paths connecting each pair.

Example: Cohesion & Clustering

Structural Cohesion

28 of 78

Node connectivity =/= density or edge connectivity!

Example: Cohesion & Clustering

Structural Cohesion

29 of 78

Serendipity

30 of 78

1

2

3

4

Nestedness Structure

Cohesive Blocks

Depth

Sociogram

5

1

2

3

4

5

6

7

8

9

Cohesive Blocking

The arrangement of subsequently more connected sets by branches and depth uniquely characterize the connectivity structure of a network

Example: Cohesion & Clustering

Structural Cohesion

31 of 78

School Attachment

Example: Cohesion & Clustering

Structural Cohesion

32 of 78

Business Political Action

Measuring Networks: Connectivity

Redundancy Global: Structural Cohesion

33 of 78

34 of 78

Hybrid Model of Firm Ownership

35 of 78

Hybrid Model of Firm Ownership

36 of 78

Hybrid Model of Firm Ownership

37 of 78

Hybrid Model of Firm Ownership

Business groups nested within the core

38 of 78

Hybrid Model of Firm Ownership

Core Members are multiply connected & higher in revenue

39 of 78

Hard Problems

40 of 78

Example: Dynamic networks

41 of 78

Example: Dynamic networks

42 of 78

43 of 78

44 of 78

Contact network: Everyone, it is a connected component

Who can “A” reach?

Discussions of network effects on STD spread often speak loosely of “the network.”

There are three relevant networks that are often conflated:

Three relevant networks

Example: Dynamic networks

45 of 78

Exposure network: here, node “A” could reach up to 8 others

Who can “A” reach?

Discussions of network effects on STD spread often speak loosely of “the network.”

There are three relevant networks that are often conflated:

Three relevant networks

Example: Dynamic networks

46 of 78

Transmission network: upper limit is 8 through the exposure links (dark blue). Transmission is path dependent: if no transmission to B, then also none to {K,L,O,J,M}

Who can “A” reach?

Exposable Link (from A’s p.o.v.)

Contact

Discussions of network effects on STD spread often speak loosely of “the network.”

There are three relevant networks that are often conflated:

Three relevant networks

Example: Dynamic networks

47 of 78

Three relevant networks

Discussions of network effects on STD spread often speak loosely of “the network.”

There are three relevant networks that are often conflated:

  1. The contact network. The set of pairs of people connected by sexual contact. G(V,E).

  • The exposure network. A subset of the edges in the contact network where timing makes it possible for one person to pass infection to another.

  • The transmission network. The subset of the exposure network where disease is actually passed. In most cases this is a tree layered on (2) and rooted on a source/seed node.

Example: Dynamic networks

48 of 78

Example: Dynamic networks

49 of 78

50 of 78

A B C D E F

A 0 1 2 2 4 1

B 1 0 1 2 3 2

C 0 1 0 1 2 2

D 0 0 1 0 1 1

E 0 0 0 1 0 2

F 1 0 0 1 0 0

While a is 2 steps from d, and d is 1 step from e, a and e are 4 steps apart.

This is because the shorter path from a to e emerges after the path from d to e ended.

4

2

1

Path distances no longer simply add

Example: Dynamic networks

51 of 78

Timing constrains potential diffusion paths in networks, since bits can flow through edges that have ended.

This means that:

    • Structural paths are not equivalent to the diffusion-relevant path set.
    • Network distances don’t build on each other.
    • Weakly connected components overlap without diffusion reaching across sets.
    • Small changes in edge timing can have dramatic effects on overall diffusion
    • Diffusion potential is maximized when edges are concurrent and minimized when they are “inter-woven” to limit reachability.

Combined, this means that many of our standard path-based network measures will be incorrect on dynamic graphs.

Example: Dynamic networks

52 of 78

The distribution of paths is important for many of the measures we typically construct on networks, and these will be change if timing is taken into consideration:

Centrality:

Closeness centrality

Path Centrality

Information Centrality

Betweenness centrality

Network Topography

Clustering

Path Distance

Groups & Roles:

Correspondence between degree-based position and reach-based position

Structural Cohesion & Embeddedness

Opportunities for Time-based block-models (similar reachability profiles)

In general, any measures that take the systems nature of the graph into account will differ.

Structural measurement implications

Example: Dynamic networks

53 of 78

New versions of classic reachability measures:

  1. Temporal reach: The ij cell = 1 if i can reach j through time.
  2. Temporal geodesic: The ij cell equals the number of steps in the shortest path linking i to j over time.
  3. Temporal paths: The ij cell equals the number of time-ordered paths linking i to j.

These will only equal the standard versions when all ties are concurrent.

Duration explicit measures

4) Quickest path: The ij cell equals the shortest time within which i could reach j.

5) Earliest path: The ij cell equals the real-clock time when i could first reach j.

6) Latest path: The ij cell equals the real-clock time when i could last reach j.

7) Exposure duration: The ij cell equals the longest (shortest) interval of time over which i could transfer a good to j.

Each of these also imply different types of “betweenness” roles for nodes or edges, such as a “limiting time” edge, which would be the edge whose comparatively short duration places the greatest limits on other paths.

Structural measurement implications

Example: Dynamic networks

54 of 78

Define time-dependent closeness as the inverse of the sum of the distances needed for an actor to reach others in the network.*

Actors with high time-dependent closeness centrality are those that can reach others in few steps. Note this is directed. Since Dij =/= Dji (in most cases) once you take time into account.

*If i cannot reach j, I set the distance to n+1

Edge timing constraints on diffusion

When is a network?

55 of 78

Define fastness centrality as the average of the clock-time needed for an actor to reach others in the network:

Actors with high fastness centrality are those that would reach the most people early. These are likely important for any “first mover” problem.

Edge timing constraints on diffusion

When is a network?

56 of 78

Define quickness centrality as the average of the minimum amount of time needed for an actor to reach others in the network:

Where Tjit is the time that j receives the good sent by i at time t, and Tit is the time that i sent the good. This then represents the shortest duration between transmission and receipt between i and j.

Note that this is a time-dependent feature, depending on when i “transmits” the good out into the population. Note min is one of many functions, since the time-to-target speed is really a profile over the duration of t.

Edge timing constraints on diffusion

When is a network?

57 of 78

Define exposure centrality as the average of the amount of time that actor j is at risk to a good introduced by actor i.

Where Tijl is the last time that j could receive the good from i and Tiif is the first time that j could receive the good from i, so the difference is the interval in time when i is at risk from j.

Edge timing constraints on diffusion

When is a network?

58 of 78

How do these centrality scores compare?

Here I compare the duration-dependent measures to the standard measures on this example graph.

Based only on the structure of the ties, not the timing, the most central nodes are nodes 13, 16 and 4.

Since this is a simulation, I permute the observed time-ranges on this graph to test the general relation between the fixed and temporal measures.

Edge timing constraints on diffusion

When is a network?

59 of 78

Here I compare the duration-dependent measures to the standard measures on this example graph.

Box plots based on 500 permutations of the observed time durations. This holds constant the duration distribution and the number of edges active at any given time.

Edge timing constraints on diffusion

When is a network?

60 of 78

61 of 78

When does timing really matter?

How do structure and dynamics intersect?

62 of 78

Measures

Dependent variable: Reachability in the exposure graph. This is the proportion of pairs in the network that are reachable in time.

Exposure Graph Density

Example: When does concurrency matter?

63 of 78

Measures: Independent variables

Features of the topology. Of key interest is the level of structural cohesion.

1

2

3

4

5

6

7

8

9

1

--

1

1

1

1

1

1

1

1

2

1

--

2

2

2

1

1

1

1

3

1

2

--

2

2

1

1

1

1

4

1

2

2

--

2

1

1

1

1

5

1

2

2

2

--

1

1

1

1

6

1

1

1

1

1

--

2

2

1

7

1

1

1

1

1

2

--

2

1

8

1

1

1

1

1

2

2

--

1

9

1

1

1

1

1

1

1

1

--

Cell Value = highest k-connected component pair belongs to.

Average = 1.25

Example: When does concurrency matter?

64 of 78

1

2

3

4

5

6

7

8

9

1

--

1

1

1

1

1

1

1

1

2

1

--

3

3

3

2

2

2

1

3

1

3

--

3

3

2

2

2

1

4

1

3

3

--

3

2

2

2

1

5

1

3

3

3

--

2

2

2

1

6

1

2

2

2

2

--

2

2

1

7

1

2

2

2

2

2

--

2

1

8

1

2

2

2

2

2

2

--

1

9

1

1

1

1

1

1

1

1

--

Cell Value = highest k-connected component pair belongs to.

Average = 1.6

Measures: Independent variables

Features of the topology. Of key interest is the level of structural cohesion.

Example: When does concurrency matter?

65 of 78

Measures: Independent variables

Features of the topology. Of key interest is the level of structural cohesion.

Average connectivity

Example: When does concurrency matter?

66 of 78

Volume

Distance

Connectivity

Nodes: 148

Mean Deg: 6.16

Density: 0.042

Centralization: 0.187

Nodes: 80

Mean Deg: 5.27

Density: 0.067

Centralization: 0.373

Nodes: 154

Mean Deg: 3.71

Density: 0.025

Centralization: 0.147

Nodes: 128

Mean Deg: 3.39

Density: 0.027

Centralization: 0.205

Mean: 3.59

Diameter: 5

Centralization: 0.312

Mean: 3.02

Diameter: 5

Centralization: 0.413

Mean: 4.99

Diameter: 8

Centralization: 0.259

Mean: 4.55

Diameter: 6

Centralization: 0.301

Largest BC: 0.51

Pairwise K: 1.57

Largest BC: 0.33

Pairwise K: 1.34

Largest BC: 0.08

Pairwise K: 1.07

Largest BC:

Pairwise K: 1.06

Exemplar independent variables

“High Cohesive”

“Low Cohesive”

Example: When does concurrency matter?

67 of 78

Example: When does concurrency matter?

68 of 78

“Low Cohesive”

Example: When does concurrency matter?

69 of 78

Proportion of relations concurrent

Density of the Exposure Network

Example: When does concurrency matter?

70 of 78

Example: When does concurrency matter?

71 of 78

Topology & Time interact: How relational sequencing affects diffusion is conditioned by the structural patterns of relations.

Examples:

- Time limitations mean star nodes can’t interact with everyone at each time tick; effects of high degree are thus limited by schedule/availabiltiy

- If within-cluster ties are also more frequent than between cluster ties, then the effects of communities will be magnified.

-Multiple connectivity should provide routes around breaks built by temporal sequence.

Network Diffusion & Peer Influence

Structural Moderators of Timing Effects

72 of 78

Solution? Turn time into a network!

Time-Space graph representations

“Stack” a dynamic network in time, compiling all “node-time” and “edge-time” events (similar to an event-history compilation of individual level data).

Consider an example:

  1. Repeat contemporary ties at each time observation, linked by relational edges as they happen.

  • Between time slices, link nodes to later selves “identity” edges

3. Time Matters

Dynamics of affect dynamics on

73 of 78

So now we:

  1. Convert every edge to a node
  2. Draw a directed arc between edges that (a) share a node and (b) precede each other in time.

Solution? Turn time into a network!

3. Time Matters

Dynamics of affect dynamics on

74 of 78

So now we:

  1. Convert every edge to a node
  2. Draw a directed arc between edges that (a) share a node and (b) precede each other in time.

  • After the transformation, concurrent relations are easily seen as reciprocal edges in the line-graph. Becomes this:

Solution? Turn time into a network!

3. Time Matters

Dynamics of affect dynamics on

75 of 78

What’s the point?

76 of 78

77 of 78

78 of 78