34
Block Modeling Overview Social life can be described (at least in part) through social roles. To the extent that roles can be characterized by regular interaction patterns, we can summarize roles through common relational patterns. Social life as interconnected system of roles Important feature: thinking of roles as connected in a role system = social structure

6 Block Modeling

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: 6 Block Modeling

Block Modeling

Overview

Social life can be described (at least in part) through social roles.

To the extent that roles can be characterized by regular interaction patterns, we can summarize roles through common relational patterns.

Social life as interconnected system of roles

Important feature: thinking of roles as connected in a role system = social structure

Page 2: 6 Block Modeling

•Rights and obligations with respect to other people or classes of people

•Roles require a ‘role compliment’ another person who the role-occupant acts with respect to

Examples:Parent - child, Teacher - student, Lover - lover, Friend - Friend,

Husband - Wife, etc.

Nadel (Following functional anthropologists and sociologists) defines ‘logical’ types of roles, and then examines how they can be linked together.

Elements of a Role

Page 3: 6 Block Modeling

Necessary:Some roles fit together necessarily. For example, the expected

interaction patterns of “son-in-law” are implied through the joint roles of “Husband” and “Spouse-Parent”

Coincidental:Some roles tend to go together empirically, but they need not

(businessman & club member, for example).

Distinguishing the two is a matter of usefulness and judgement, but relates to social substitutability. The distinction reverts to how the system as a whole will be held together in the face of changes in role occupants.

Coherence of Role Systems

Page 4: 6 Block Modeling

Empirical social structures

• With the fall of functionalism in the late 60s, many of the ideas about social structure and system were also tossed.

• White et al demonstrate how we can understand social structure as the intercalation of roles, without the a priori logical categories.

• Empirical role is: • A set of relations signifying exchange of something (support, ideas,

commands, etc) between actors.

Page 5: 6 Block Modeling

Start with some basic ideas of what a role is: An exchange of something (support, ideas, commands, etc) between actors. Thus, we might represent a family as:

H W

C

C

C

Provides food for

Romantic Love

Bickers with(and there are, of course, many other relations inside the family)

Family Structure

Page 6: 6 Block Modeling

White et al: From logical role systems to empirical social structures

The key idea, is that we can express a role through a relation (or set of relations) and thus a social system by the inventory of roles. If roles equate to positions in an exchange system, then we need only identify particular aspects of a position. But what aspect?

Structural Equivalence•Two actors are structurally equivalent if they have the same types of ties to the same people.

Generalization

Page 7: 6 Block Modeling

Structural Equivalence

A single relation

Page 8: 6 Block Modeling

Structural Equivalence

Graph reduced to positions

Page 9: 6 Block Modeling

Alternative notions of equivalence

Instead of exact same ties to exact same alters, you look for nodes with similar ties to similar types of alters

Page 10: 6 Block Modeling

In any positional analysis, there are 4 basic steps:

1) Identify a definition of equivalence2) Measure the degree to which pairs of actors are equivalent3) Develop a representation of the equivalencies4) Assess the adequacy of the representation5) Repeat and refine

Basic Steps: Blockmodeling

Page 11: 6 Block Modeling

1) Identify a definition of equivalence

Structural Equivalence: Two actors are equivalent if they have the same type of ties to the same people.

Page 12: 6 Block Modeling

AutoMorphic Equivalence

• Actors occupy indistinguishable structural locations in the network. That is, that they are in isomorphic positions in the network.

• Two graphs are isomorphic if there is some mapping of nodes to positions that equates the two.

• In general, automorphically equivalent nodes are equivalent with respect to all graph theoretic properties (I.e. degree, number of people reachable, centrality, etc.)

Page 13: 6 Block Modeling

Automorphic Equivalence:

Page 14: 6 Block Modeling

Regular equivalence does not require actors to have identical ties to identical actors or to be structurally indistinguishable.

Actors who are regularly equivalent have identical ties to and from equivalent actors.

If actors i and j are regularly equivalent, then for all relations and for all actors, if i --> k, then there exists some actor l such that j--> l and k is regularly equivalent to l.

Regular Equivalence

i j

k l

Page 15: 6 Block Modeling

Regular Equivalence:

There may be multiple regular equivalence partitions in a network, and thus we tend to want to find the maximal regular equivalence position, the one with the fewest positions.

Page 16: 6 Block Modeling

Note that:Structurally equivalent actors are automorphically equivalent,Automorphically equivalent actors are regularly equivalent.

Structurally equivalent and automorphically equivalent actors are role equivalent

In practice, we tend to ignore some of these fine distinctions, as they get blurred quickly once we have to operationalize them in real graphs. It turns out that few people are ever exactly equivalent, and thus we approximate the links between the types.

In all cases, the procedure can work over multiple relations simultaneously.

The process of identifying positions is called blockmodeling, and requires identifying a measure of similarity among nodes.

Practicality

Page 17: 6 Block Modeling

0 1 1 1 0 0 0 0 0 0 0 0 0 01 0 0 0 1 1 0 0 0 0 0 0 0 01 0 0 1 0 0 1 1 1 1 0 0 0 01 0 1 0 0 0 1 1 1 1 0 0 0 00 1 0 0 0 1 0 0 0 0 1 1 1 10 1 0 0 1 0 0 0 0 0 1 1 1 10 0 1 1 0 0 0 0 0 0 0 0 0 00 0 1 1 0 0 0 0 0 0 0 0 0 00 0 1 1 0 0 0 0 0 0 0 0 0 00 0 1 1 0 0 0 0 0 0 0 0 0 00 0 0 0 1 1 0 0 0 0 0 0 0 00 0 0 0 1 1 0 0 0 0 0 0 0 00 0 0 0 1 1 0 0 0 0 0 0 0 00 0 0 0 1 1 0 0 0 0 0 0 0 0

Blockmodeling is the process of identifying these types of positions. A block is a section of the adjacency matrix - a “group” of people.

Here I have blocked structurally equivalent actors

Page 18: 6 Block Modeling

. 1 1 1 0 0 0 0 0 0 0 0 0 01 . 0 0 1 1 0 0 0 0 0 0 0 01 0 . 1 0 0 1 1 1 1 0 0 0 01 0 1 . 0 0 1 1 1 1 0 0 0 00 1 0 0 . 1 0 0 0 0 1 1 1 10 1 0 0 1 . 0 0 0 0 1 1 1 10 0 1 1 0 0 . 0 0 0 0 0 0 00 0 1 1 0 0 0 . 0 0 0 0 0 00 0 1 1 0 0 0 0 . 0 0 0 0 00 0 1 1 0 0 0 0 0 . 0 0 0 00 0 0 0 1 1 0 0 0 0 . 0 0 00 0 0 0 1 1 0 0 0 0 0 . 0 00 0 0 0 1 1 0 0 0 0 0 0 . 00 0 0 0 1 1 0 0 0 0 0 0 0 .

1 2 3 4 5 61 0 1 1 0 0 02 1 0 0 1 0 03 1 0 1 0 1 04 0 1 0 1 0 1 5 0 0 1 0 0 06 0 0 0 1 0 0

Once you block the matrix, reduce it, based on the number of ties in the cell of interest. The key values are a zero block (no ties) and a one-block (all ties present):

Structural equivalence thus generates 6 positions in the network

1 2 3 4 5 6

12

3

4

5

6

Page 19: 6 Block Modeling

. 1 1 1 0 0 0 0 0 0 0 0 0 01 . 0 0 1 1 0 0 0 0 0 0 0 01 0 . 1 0 0 1 1 1 1 0 0 0 01 0 1 . 0 0 1 1 1 1 0 0 0 00 1 0 0 . 1 0 0 0 0 1 1 1 10 1 0 0 1 . 0 0 0 0 1 1 1 10 0 1 1 0 0 . 0 0 0 0 0 0 00 0 1 1 0 0 0 . 0 0 0 0 0 00 0 1 1 0 0 0 0 . 0 0 0 0 00 0 1 1 0 0 0 0 0 . 0 0 0 00 0 0 0 1 1 0 0 0 0 . 0 0 00 0 0 0 1 1 0 0 0 0 0 . 0 00 0 0 0 1 1 0 0 0 0 0 0 . 00 0 0 0 1 1 0 0 0 0 0 0 0 .

1 2 31 1 1 02 1 1 1 3 0 1 0

Once you partition the matrix, reduce it:

Regular equivalence

1 2

3

Page 20: 6 Block Modeling

To get a block model, you have to measure the similarity between each pair. If two actors are structurally equivalent, then they will have exactly similar patterns of ties to other people. Consider the example again:

. 1 1 1 0 0 0 0 0 0 0 0 0 01 . 0 0 1 1 0 0 0 0 0 0 0 01 0 . 1 0 0 1 1 1 1 0 0 0 01 0 1 . 0 0 1 1 1 1 0 0 0 00 1 0 0 . 1 0 0 0 0 1 1 1 10 1 0 0 1 . 0 0 0 0 1 1 1 10 0 1 1 0 0 . 0 0 0 0 0 0 00 0 1 1 0 0 0 . 0 0 0 0 0 00 0 1 1 0 0 0 0 . 0 0 0 0 00 0 1 1 0 0 0 0 0 . 0 0 0 00 0 0 0 1 1 0 0 0 0 . 0 0 00 0 0 0 1 1 0 0 0 0 0 . 0 00 0 0 0 1 1 0 0 0 0 0 0 . 00 0 0 0 1 1 0 0 0 0 0 0 0 .

1 2 3 4 5 6

12

3

4

5

6

C D Match1 1 10 0 1. 1 01 . 00 0 10 0 11 1 1 1 1 1 1 1 11 1 10 0 10 0 10 0 10 0 1Sum: 12

C and D match on 12 other people

Page 21: 6 Block Modeling

If the model is going to be based on asymmetric or multiple relations, you simply stack the various relations:

H W

CC

C

Provides food for

Romantic Love

Bickers with

Romance0 1 0 0 01 0 0 0 00 0 0 0 00 0 0 0 00 0 0 0 0

Feeds0 0 1 1 10 0 1 1 10 0 0 0 00 0 0 0 00 0 0 0 0

Bicker0 0 0 0 00 0 0 0 00 0 0 1 10 0 1 0 00 0 1 1 0

0 1 0 0 01 0 0 0 00 0 0 0 00 0 0 0 00 0 0 0 00 0 1 1 10 0 1 1 10 0 0 0 00 0 0 0 00 0 0 0 00 0 0 0 00 0 0 0 01 1 0 0 01 1 0 0 01 1 0 0 00 0 0 0 00 0 0 0 00 0 0 1 10 0 1 0 10 0 1 1 0

Stacked

Page 22: 6 Block Modeling

0 8 7 7 5 5 11 11 11 11 7 7 7 7 8 0 5 5 7 7 7 7 7 7 11 11 11 11 7 5 0 12 0 0 8 8 8 8 4 4 4 4 7 5 12 0 0 0 8 8 8 8 4 4 4 4 5 7 0 0 0 12 4 4 4 4 8 8 8 8 5 7 0 0 12 0 4 4 4 4 8 8 8 811 7 8 8 4 4 0 12 12 12 8 8 8 811 7 8 8 4 4 12 0 12 12 8 8 8 811 7 8 8 4 4 12 12 0 12 8 8 8 811 7 8 8 4 4 12 12 12 0 8 8 8 8 7 11 4 4 8 8 8 8 8 8 0 12 12 12 7 11 4 4 8 8 8 8 8 8 12 0 12 12 7 11 4 4 8 8 8 8 8 8 12 12 0 12 7 11 4 4 8 8 8 8 8 8 12 12 12 0

For the entire matrix, we get:

(number of agreements for each ij pair)

Page 23: 6 Block Modeling

1.00 -0.20 0.08 0.08 -0.19 -0.19 0.77 0.77 0.77 0.77 -0.26 -0.26 -0.26 -0.26-0.20 1.00 -0.19 -0.19 0.08 0.08 -0.26 -0.26 -0.26 -0.26 0.77 0.77 0.77 0.77 0.08 -0.19 1.00 1.00 -1.00 -1.00 0.36 0.36 0.36 0.36 -0.45 -0.45 -0.45 -0.45 0.08 -0.19 1.00 1.00 -1.00 -1.00 0.36 0.36 0.36 0.36 -0.45 -0.45 -0.45 -0.45-0.19 0.08 -1.00 -1.00 1.00 1.00 -0.45 -0.45 -0.45 -0.45 0.36 0.36 0.36 0.36-0.19 0.08 -1.00 -1.00 1.00 1.00 -0.45 -0.45 -0.45 -0.45 0.36 0.36 0.36 0.36 0.77 -0.26 0.36 0.36 -0.45 -0.45 1.00 1.00 1.00 1.00 -0.20 -0.20 -0.20 -0.20 0.77 -0.26 0.36 0.36 -0.45 -0.45 1.00 1.00 1.00 1.00 -0.20 -0.20 -0.20 -0.20 0.77 -0.26 0.36 0.36 -0.45 -0.45 1.00 1.00 1.00 1.00 -0.20 -0.20 -0.20 -0.20 0.77 -0.26 0.36 0.36 -0.45 -0.45 1.00 1.00 1.00 1.00 -0.20 -0.20 -0.20 -0.20-0.26 0.77 -0.45 -0.45 0.36 0.36 -0.20 -0.20 -0.20 -0.20 1.00 1.00 1.00 1.00-0.26 0.77 -0.45 -0.45 0.36 0.36 -0.20 -0.20 -0.20 -0.20 1.00 1.00 1.00 1.00-0.26 0.77 -0.45 -0.45 0.36 0.36 -0.20 -0.20 -0.20 -0.20 1.00 1.00 1.00 1.00-0.26 0.77 -0.45 -0.45 0.36 0.36 -0.20 -0.20 -0.20 -0.20 1.00 1.00 1.00 1.00

Correlation between each node’s set of ties.

For the example, this would be:

Measuring similarity

Page 24: 6 Block Modeling

The initial method for finding structurally equivalent positions was CONCOR, the CONvergence of iterated CORrelations.

1.00 -.77 0.55 0.55 -.57 -.57 0.95 0.95 0.95 0.95 -.75 -.75 -.75 -.75-.77 1.00 -.57 -.57 0.55 0.55 -.75 -.75 -.75 -.75 0.95 0.95 0.95 0.950.55 -.57 1.00 1.00 -1.0 -1.0 0.73 0.73 0.73 0.73 -.75 -.75 -.75 -.750.55 -.57 1.00 1.00 -1.0 -1.0 0.73 0.73 0.73 0.73 -.75 -.75 -.75 -.75-.57 0.55 -1.0 -1.0 1.00 1.00 -.75 -.75 -.75 -.75 0.73 0.73 0.73 0.73-.57 0.55 -1.0 -1.0 1.00 1.00 -.75 -.75 -.75 -.75 0.73 0.73 0.73 0.730.95 -.75 0.73 0.73 -.75 -.75 1.00 1.00 1.00 1.00 -.77 -.77 -.77 -.770.95 -.75 0.73 0.73 -.75 -.75 1.00 1.00 1.00 1.00 -.77 -.77 -.77 -.770.95 -.75 0.73 0.73 -.75 -.75 1.00 1.00 1.00 1.00 -.77 -.77 -.77 -.770.95 -.75 0.73 0.73 -.75 -.75 1.00 1.00 1.00 1.00 -.77 -.77 -.77 -.77-.75 0.95 -.75 -.75 0.73 0.73 -.77 -.77 -.77 -.77 1.00 1.00 1.00 1.00-.75 0.95 -.75 -.75 0.73 0.73 -.77 -.77 -.77 -.77 1.00 1.00 1.00 1.00-.75 0.95 -.75 -.75 0.73 0.73 -.77 -.77 -.77 -.77 1.00 1.00 1.00 1.00-.75 0.95 -.75 -.75 0.73 0.73 -.77 -.77 -.77 -.77 1.00 1.00 1.00 1.00

Concor iteration 1:

Page 25: 6 Block Modeling

Concor iteration 2:1.00 -.99 0.94 0.94 -.94 -.94 0.99 0.99 0.99 0.99 -.99 -.99 -.99 -.99-.99 1.00 -.94 -.94 0.94 0.94 -.99 -.99 -.99 -.99 0.99 0.99 0.99 0.990.94 -.94 1.00 1.00 -1.0 -1.0 0.97 0.97 0.97 0.97 -.97 -.97 -.97 -.970.94 -.94 1.00 1.00 -1.0 -1.0 0.97 0.97 0.97 0.97 -.97 -.97 -.97 -.97-.94 0.94 -1.0 -1.0 1.00 1.00 -.97 -.97 -.97 -.97 0.97 0.97 0.97 0.97-.94 0.94 -1.0 -1.0 1.00 1.00 -.97 -.97 -.97 -.97 0.97 0.97 0.97 0.970.99 -.99 0.97 0.97 -.97 -.97 1.00 1.00 1.00 1.00 -.99 -.99 -.99 -.990.99 -.99 0.97 0.97 -.97 -.97 1.00 1.00 1.00 1.00 -.99 -.99 -.99 -.990.99 -.99 0.97 0.97 -.97 -.97 1.00 1.00 1.00 1.00 -.99 -.99 -.99 -.990.99 -.99 0.97 0.97 -.97 -.97 1.00 1.00 1.00 1.00 -.99 -.99 -.99 -.99-.99 0.99 -.97 -.97 0.97 0.97 -.99 -.99 -.99 -.99 1.00 1.00 1.00 1.00-.99 0.99 -.97 -.97 0.97 0.97 -.99 -.99 -.99 -.99 1.00 1.00 1.00 1.00-.99 0.99 -.97 -.97 0.97 0.97 -.99 -.99 -.99 -.99 1.00 1.00 1.00 1.00-.99 0.99 -.97 -.97 0.97 0.97 -.99 -.99 -.99 -.99 1.00 1.00 1.00 1.00

The initial method for finding structurally equivalent positions was CONCOR, the CONvergence of iterated CORrelations.

Page 26: 6 Block Modeling

Padget and Ansell:“Robust Action and the Rise of the Medici”

•Substantive question relates to effective state-building: there is a tension between the need to control and organization and the ability to build the legitimacy and recognition required for reproduction. The distinction between “boss” and “judge”

•They use the marriage, economic and patronage networks

•Empirically, we know that the state oligarchy structure of Florence stabilized after the rise of the medici:

Page 27: 6 Block Modeling

Padget and Ansell:“Robust Action and the Rise of the Medici”

Medici Takeover

Page 28: 6 Block Modeling

Padget and Ansell:“Robust Action and the Rise of the Medici”

The story they tell revolves around how Cosimo de’Medici was able to found a system that lasted nearly 300 years, uniting a fractured political structure.

The paradox of Cosimo is that he didn’t seem to fit the role of a Machiavellian leader as decisive and goal oriented.

The answer lies in the power resulting from ‘robust action’ embedded in a network of relations that gives rise to no clear meaning and obligation, but instead allows for multiple meanings and obligations.

Page 29: 6 Block Modeling

A real example:Padget and Ansell:“Robust Action and the Rise of the Medici”

“Political Groups” in the attribute sense do not seem to exist, so P&A turn to the pattern of network relations among families.

This is the BLOCK reduction of the full 92 family network.

Page 30: 6 Block Modeling

An example:Relations among Italian families.

Political and friendship ties

Page 31: 6 Block Modeling

Generalized Block Models

The recent work on generalization focuses on the patterns that determine a block.

Instead of focusing on just the density of a block, you can identify a block as any set that has a particular pattern of ties to any other set. Examples include:

Page 32: 6 Block Modeling

Generalized Block Models

Page 33: 6 Block Modeling

Compound Relations.

One of the most powerful tools in role analysis involves looking at role systems through compound relations.

A compound relation is formed by combining relations in single dimensions. The best example of compound relations come from kinship.

SiblingChild of

Sibling0 1 0 0 01 0 0 0 00 0 0 1 00 0 1 0 00 0 0 0 0

Child of0 0 1 1 00 0 0 0 10 0 0 0 00 0 0 0 00 0 0 0 0

x =

Nephew/Niece0 0 0 0 10 0 1 1 00 0 0 0 00 0 0 0 00 0 0 0 0

SC = SC

Page 34: 6 Block Modeling

An example of compound relations can be found in W&F. This role table catalogues the compounds for two relations “Is boss of” and “Is on the same level as”