Introduction to Social Network Methods: Chapter 17: Two-mode networks

Literature › Works

Robert A. Hanneman; Mark Riddle (2005). Introduction to Social Network Methods: Chapter 17: Two-mode networks. Introduction to Social Network Methods (on-line text, University of California, Riverside). University of California, Riverside.

Type	book chapter
Scope	Comparative and methods
Method	theoretical or conceptual
Period studied	2000–2004
Regions	California (illustrative data only)
Themes	methods data sources, comparative theory
Citation key	hanneman2005

Argument

The chapter introduces methods for analysing two-mode (actor-by-event, or affiliation) network data. It argues that such data help reveal macro-micro and structure-agency relations, because actors define events through participation while events shape the actors' choices. It shows how to convert two-mode data to one-mode matrices, how to visualise it as a bipartite graph, and how to find patterns with scaling methods (SVD, factor analysis, correspondence analysis) and block models (core-periphery, factions). The techniques are illustrated with data on large donors to California ballot initiatives. It has no content on South Africa or apartheid.

Evidence and method

Illustrative data sets only: Davis et al.'s Deep South data (18 women at 14 events) and a donors-by-initiatives matrix (23 donors by 44 California ballot initiatives, 2000-2004), analysed with UCINET and NetDraw.

Findings

• Two-mode data link actors and events and can be converted into two one-mode matrices. (p. 3)

  A very common and very useful approach to two-mode data is to convert it into two one-mode data sets, and examine relations within each mode separately.

• A bipartite matrix places actors and events in one square matrix, which must be interpreted with caution. (p. 5)

  Considerable caution is needed in interpretation, because the network that is being analyzed is a very unusual one in which the relations are ties between nodes at different levels of analysis.

• The SVD of the donor data shows no simple low-dimensional structure. (p. 8)

  This result tells us that the ways that actors and events "go together" is not clean, simple, and easy -- in this case.

• Factor analysis and SVD are not recommended for binary data. (p. 14)

  For binary data, the use of factor analysis and SVD is not recommended.

• In the SVD results, Democrats and unions cluster together on one dimension, opposite Republicans and business groups. (p. 10)

  Toward the positive end of dimension one (which we earlier interpreted as favoring public expenditure) we find the Democratic party, public employees and teachers unions; at the opposite pole, we find Republicans and some business and professional groups.

• Scaling binary data is risky and block modelling offers an alternative. (p. 17)

  Block modeling works directly on the binary incidence matrix by trying to permute rows and columns to fit, as closely as possible, idealized images.

• The core-periphery model fits the donor data reasonably well. (p. 19)

  While far from perfect, the model here is good enough to be taken seriously.

• The two-faction model fits worse than core-periphery. (p. 21)

  The fit of the two factions model is not as impressive as the fit of the core-periphery model.