Essex Summer School in
Social Science Data Analysis

Filip Agneessens

Filip Agneessens is a Researcher at the Department of Sociology University of Groningen and the Interuniversity Center for Social Science Theory and Methodology (ICS). He has published a number of papers on social network analysis in leading social science journals. Together with Martin Everett he is a guest-editor for a special issue on two-mode social network analysis in the journal Social Networks. He has previously taught a number of courses on network theory, social network methods and social network analysis applied to organisations, sociology and political sciences. His research centers on network formation within organisations and their impact on attitudes and behaviour of individual employees. He has also been working on the development of mathematical models for analyzing networks, on imputation techniques for solving missing data problems and on methods for building social support typologies for ego-networks. In recent work he has applied exponential random graph models (p* models) to study cultural participation, inter-organisational networks, and (co)sponsorship among US senators.

Advanced Social Network Analysis II – Longitudinal Network Analysis for Social Selection and Social Influence Mechanisms

Filip Agneessens, University of Groningen
5 - 16 August (two week course / 35 hrs)

Detailed Course Outline [PDF]

Course Content

This course focuses on different longitudinal methods for analyzing social network data. We focus both on social selection and social influence mechanisms, and discuss different statistical techniques for exploring and testing such hypotheses (with longitudinal network data). The course will enable students to answer questions related to social selection, such as: “Is an advice relation more likely to emerge among people who are similar? (homophily)” or “Do friends of friends tend to be friends? (closure)”. The methods discussed will also be able to answer questions on social influence (and contagion), such as: “Do people who have more advice ties tend to be better performing?” or “Do adolescents start smoking if their friends smoke?”

While Advanced Social Network Analysis I focuses on testing (selection) mechanisms with cross-sectional network data, this course focuses on longitudinal methods and therefore requires network data at different time points. Longitudinal network data can either be: 1) network data collected at specific time points (e.g., asking students (via surveys) about their friendship relations and their smoking behavior at the beginning of the academic year and again 3 months later), or 2) data that represent relational “events” (e.g., data on who emailed whom, or who called whom, and the exact time at which this happened).

The course starts with a discussion of some relatively simple techniques for looking at longitudinal network data (measures of change and visualization techniques to visualize these changes). We then move onto the analysis of changes in people’s ego-networks over time (churn). We subsequently turn to more advanced longitudinal models to test selection and the co-evolution of both selection and influence mechanisms simultaneously using stochastic actor-based model (RSiena), as well as event based models. (See the detailed course description for more details about the course) Software used includes: Visone, the packages “sna”, “spdep”, “RSiena” and “relevant” in R. .

Course Objectives

The course aims to familiarize participants with statistical methods for longitudinal social network data. Both methods for the analysis of network relations being measured at specific time points (e.g., friendship measured via a survey at three different time points), and relational “events” being observed at specific moments in time (such as emails being send) will be discussed in the course. In addition both social selection and social influence processes will be dealt with. Participants will become familiar with specific programs designed for these analyses. The course focuses on: understanding the mathematical basis for the modeling, conducting statistical analyses, making appropriate model selection (including goodness of fit), interpreting the output and linking the different parameters to different (network) theories. Participants are encouraged to bring with them their own network data to be analyzed using the techniques covered.

Participants should be able to answer questions, such as:

• Are actors who are more central more likely to have a high performance?
• Do employees benefit from building new (brokerage) ties?
• Are pupils influenced in their political views, music taste or smoking behaviour by the political views, music taste or smoking behaviour of their friends?
• Do employees trust others more if they have a similar attitude towards work, or do their attitudes change as a result of their position in the trust network?
• Does having criminal friends make students more likely to become criminal as well, or do students who have a similar tendency to criminal behaviour tend to develop friendships?

Course Prerequisites

Participants should have taken a basic course in social network analysis, so be familiar with such terms as mutuality, transitivity, degree, centrality. Participants should also have taken a basic course in (logistic) regression analysis..

Reading

Carrington, P.J., J. Scott, and S. Wasserman. 2005.Models and Methods in Social Network Analysis.Cambridge University Press.

Scott, J. 1992. Social Network Analysis. Sage.

Wasserman, S., and Faust, K. 1996. Social Network Analysis. Cambridge University Press.

[top of page]