Social Network Analysis for Newbies (Tuesday, September 4, 2012, morning)

Thomas N. Friemel, University of Zurich

Recent developments in our society made apparent how important it is to understand social networks. Even though social networks existed ever since and social network analysis (SNA) as a scientific discip-line dates back several decades we currently face a major trend to apply SNA to an increasing number of topics and research questions.

This introduction to SNA requires no prior knowledge. In a first step it provides answers to basic ques-tions as what network analysis is (definition), where it comes from (historical development), and why it is of any importance (relevance). The introduction to important terms and concepts provides an insight into different areas of applications and central research questions. Various techniques of data collection, data storage, and simple network measures are presented. Different computer programs will be briefly dis-cussed to illustrate which programs are suitable for which kind of data and analysis.

Applications of Social Network Analysis (Tuesday, September 4, 2012, afternoon)

Filip Agneessens, University of Groningen
Thomas N. Friemel, University of Zurich
Manuel Fischer, University of Geneva
Karin Ingold, University of Bern

This course on Applications of Social Network analysis provides insight in the historic development and latest research conducted in political science, mass communication, and sociology. In the first part of this seminar-style course we will focus on the applications in the domain of political science. Topics such as policy networks, governance and regulatory networks, and the use of SNA for the analysis of coalitions, power, and political decision-making will be addressed. We will discuss how the network concept and the method of SNA can provide a useful tool for such analyses at the level of policy domains, countries, or the supra-national level. In the second part on mass communication & media research (slides as a pdf for download) a focus will be on different roles in communication networks (e.g. opinion leaders vs. followers) which can be found on the micro level of individual, the meso level or organizations, as well as the macro level of global communication structure. The third part discusses central SNA applications in sociology which include classic aspects like “the strength of weak ties”, “small world” phenomena, and theory and measurement of individual, group and network level “social capital”, as well as up to date applications in organizational sociology.

Some basic knowledge of SNA (provided in the introductory course in the morning) are helpful to fully profit from this course. It is designed for doctoral students and researchers who consider applying SNA in their work.

The Analysis of Longitudinal Social Network Data (Tuesday, September 4, 2012, full day)

Johan Koskinen, University of Manchester

The workshop will give an introduction to statistical modeling of longitudinal network data and demon-strate the basics of using the RSiena program. The first part of the workshop will try to develop an under-standing of stochastic actor-oriented models from the joint perspectives of the underlying statistical me-thodology and practical issues of data analysis. A first objective is that the participant will be able to read in data, run the model, and interpreting the results as well as check for the adequacy of the analysis. The most basic model – for the evolution of single directed graphs - will then be extended to models for the joint analysis of the evolution of network ties and behaviour.
The second part will continue to cover additional issues in longitudinal analysis. Firstly different model specifications, model checks and alternative methods of estimation, secondly models for other types of network data. Examples of the latter include bipartite data, the evolution of multiple networks, multilevel networks and undirected networks.

No previous experience of using R is required or necessary for the workshop. In order to follow the prac-tical examples, the participant should however bring a laptop with R and RSiena installed. The following preparations may be of help:
• A laptop with the R environment installed (www.r-project.org/)
• An acquaintance with actor oriented models (a good reference is
www.stats.ox.ac.uk/~snijders/siena/SnijdersSteglichVdBunt2009.pdf, or
for the more mathematically inclined
www.stats.ox.ac.uk/~snijders/siena/Longi_Net.pdf)
Optional preparation which will maximize the attendee's experience
includes the following:
• Chapter two of the RSiena manual (Manual for Siena 4.0 )
• Sessions 1–3 of Ruth Ripley's R course
www.stats.ox.ac.uk/~ruth/RCourse/oxford.html
• Investigation of the RSiena webpage www.stats.ox.ac.uk/~snijders/siena/

ERGMs for Social Networks (Wednesday, September 5, 2012, morning) 

Skyler Cranmer, University of Carolina at Chapel Hill

This workshop introduces exponential random graph models as a means of making statistical inferences on social/political/policy networks. The workshop consists of three basic parts. First, we will discuss the problem of inference on networks: we will very briefly cover the anatomy of networks, then discuss why classical regression models rarely produce unbiased estimates on network data. Second, we will introduce the ERGM both mathematically (med/low mathematical detail) and substantively, considering its estimation, interpretation, and limitations. Third we will consider extensions of the ERGM to longitudinal networks, valued-edge networks, and using the ERGM as a forecasting tool. A few examples of how to estimate ERGMs in R will be covered throughout.

No experience with the ERGM is assumed. Participants will be expected to understand fundamental social network concepts and terminology and have some background in standard statistical procedures (e.g. regression and logistic regression). No experience with R is required, though it may make the examples a bit easier. 

visone - Analysis and Visualization of Social Networks (Wednesday, September 5, 2012, morning)

Ulrik Brandes, Uwe Nagel and Martin Mader, University of Konstanz

This is a hands-on introduction to visone, a graphically oriented software tool that combines comprehensive means for analysis with unique visualization capabilities. After a brief introduction to its design and features, we will explore some of the core functionality of visone using exemplary network analyses; step-by-step from data input to presentation of results. We will also discuss advanced features for longitudinal network visualization, our graphical support for modeling with RSiena, and the integration of visone with R.

Some elementary knowledge of social network analysis is required for this workshop, and it is advisable to bring a laptop running Windows, MacOS, or Linux. visone (ital.: mink) is written in Java and freely available from www.visone.info. It features many standard and non-standard methods for analysis and visualization of networks, and a powerful graphical user interface. It's native file format is GraphML, allowing for arbitrarily many attributes of nodes, links, and networks, but other formats such as CSV tables, UCINet DL, Pajek .net, etc., can be imported. Visualizations can be exported as pdf, png, tiff, svg, or Windows metafiles.

Pajek XXL (cancelled, replaced by Block-modelling, see below)

Andrej Mrvar, University of Ljubljana

Pajek is program for analysis and visualization of large networks. It has been developed since 1996.  Pajek is now Windows 32 and 64 program with Unicode support. It is free for a non-commercial use: pajek.imfm.si

Since Spring 2012 a special edition of program - PajekXXL - is available: mrvar.fdv.uni-lj.si/pajek/

PajekXXL has much lower memory consumption.  For the same network it needs at least 2-3 times less physical memory than Pajek.  Operations that are memory intensive (e.g. generating random networks, extracting, shrinking,...) are also faster. Sparse networks having up to 100 millions of vertices can be analysed on computers with 4 GRAM memory. Sparse networks having up to a billion of vertices can be analysed with 16 GRAM (or more) using PajekXXL.

During the workshop we will explain what can (and what cannot) be done when networks have some 100s of millions of vertices. Typical steps how to analyse huge networks with Pajek-XXL, and how to store results in the form that can be further analysed (and visualized) with 'ordinary' Pajek will be presented.

Workshop will be based on the textbook:
de Nooy, W., Mrvar, A., and Batagelj, V. (2011): Exploratory Social Network Analysis with Pajek: Revised and Expanded Second Edition. New York: Cambridge University Press.  www.cambridge.org/9780521174800

Although we will begin workshop with a brief explanation of basic 'ordinary' Pajek usage participants are expected to be familiar at least with basic Pajek usage as covered in Chapters 1, 2, and 13 of the workshop textbook.

Block-Modelling (Wednesday, September 5, 2012, afternoon)

Filip Agneessens, University of Groningen

Filip volunteered to jump in for Andrej Mrvar and is giving a workshop on block-modelling. He is going to illustrate this SNA technique with the help of UCINET and Pajek.

Please come to the workshop with the software installed.

 

Discourse Network Analysis (Wednesday, September 5, 2012, afternoon)

Philip Leifeld, Eawag and University of Bern

Discourse is regarded as an important phenomenon in many disciplines in the social sciences and humanities. Social network analysis and qualitative content analysis can be combined in order to analyse how persons or organisations choose discursive concepts, how actors are interrelated by common arguments or solution concepts for problems, and how ideologies or frames emerge from these empirically observable patterns.

This workshop introduces the software Discourse Network Analyzer, which was developed to code text data like newspaper articles (similar to other QDA packages) and then convert these structured data into various kinds of networks that can be imported into visone, Ucinet and R. The session covers the basic theoretical and methodological concepts, examples from several real-world policy debates, a hands-on tutorial for the software, and some dynamic extensions of the basic models.

No previous experience with network analysis or content analysis is required. However, participants who want to follow the tutorial part on their own laptops should make sure that a recent version of Java is installed and download the software and sample file from http://www.philipleifeld.de/discourse-network-analyzer/