"Introduction to Social Network Analysis"
Introduction to Social Network Analysis 2009 Spring School on Transition Studies in Tartu Tetiana Kostiuchenko PhD candidate (Aspirantura) National University of “Kyiv-Mohyla Academy” Ukraine Agenda Background of social network analysis (SNA) Basic terms and concepts Centrality measures Visualization of network Why to apply in transition / transformation studies? Because SNA approach allow not only to draw a ‘map of ties’ between structures/ institutions/ societies as macro-level of social reality… … but, in addition, to track changes that occur with these ‘actors’ during social transformations in the contexts of interconnections at the micro-level (meaning connections between individuals, households or organizations) Source: Knoke & Kuklinski “Network Analysis” (1982) What were the starting points? Sociological and economic theories (Social exchange theory, social and human capital) Empirical researches (anthropology, psychology, corporate structure, markets) SNA Mathematical modeling (graph theory, systems theory, neural networks, etc.) Classical structural approach vs. SNA Classical structural SNA => approach => relations between attributes of actors actors “Attributes are intrinsic “Relation is …an characteristics of people, vs. emergent property of objects, or events” * the connection or linkage between units of observation” * Source: Knoke & Kuklinski “Network Analysis” (1982) Social network is... ... Model of material and non-material resources exchange in particular social group, community; ... ‘horizontal’ structure with interacting elements; (in contrast to the formal hierarchical structure of organization); ... OR other empirical definition that comes from particular research needs, objects, units of analysis, researcher`s background, etc. Methodological background: mathematical modeling • Graph theory Leonard Euler (1736): Seven Bridges of Königsberg B C D Methodological background: graph theory • Graph is a collection of vertices and edges between them (the most convenient form of representing any ‘structure’) Objects = vertices of the graph/ nodes Connections = edges that connect pairs of vertices Number of edges = size of the graph Methodological background: measuring relations • Sociometry (Moreno) measuring of social relationships through ‘emotional’ ties, reciprocated ‘likes’ and ‘dislikes’ within a small group Sociometric matrix Sociogram Types of relations Types of actors Transactions individuals (people) Communication organizations (information exchange) households Boundary penetration (affiliation) words Instrumental (getting a countries and job); etc... Sentiment (affection, admiration, or hostility toward each other) Authority/power (issuing and obeying commands) Kinship Relations -Different types of relations identify different networks, even when imposed on the identical set of elements; - Relations can be measured through : 1) intensity, or strength, 2) level of joint of the link between two involvement in actors the same activities (reciprocity) (i.e. volume of transacted resource; frequency of transactions) Network structure DIRECTED UNDIRECTED Bonnie Bob Biff OR Betty Betsy COMPLETE NETWORK: INCOMPLETE NETWORK: information about information about patterning of ties among patterning of ties among ALL actors some actors is absent (existence of isolates) Types of network structures CONNECTIVITY HIGH LOW LOW DOMINATION HIGH “Path”, “trail”, AND “walk” • Path: do not repeat nodes 10 ▫ 1-2-3-4-5-6-7-8 12 ▫ NOT 7-1-2-3-7-4 11 8 9 • Trail: do not repeat edges ▫ 1-2-3-1-7-8 2 7 ▫ NOT 7-1-2-7-1-4 3 • Walk: without limits 1 6 ▫ 1-2-3-1-2-7-1-7-1 4 5 Basic measures for the analysis • Size of the network (the number of existing connections) • Density (the number of ties divided by the number of pairs, times 100) • Centrality measures - degree Degree: Direct connections to other actors, higher scores are considered - closeness as better Closeness: farness from neighbors, the Path-based - betweeness lowest scores mean being closer Betweeness:«gatekeepers» who connect subgroups, clusters; higher scores means that if to remove this Walk-based => eigenvector node, the segments will probably be disconnected Eigenvector: «popularity» through being connected with well-connected Eigenvector Degree Betweeness Closeness Data courtesy of David Krackhardt (the example from UCINET software package) Interlocking Directorates in the Corporate Community by G. William Domhoff Political Elite Network: ALL types of ties Applications of SNA approach in different disciplines Sociology Anthropology Psychology Political Science SNA Ecology Epidemiology Linguistics Criminology еtc… ... and in particular: “Buzzing” in marketing Organizational network analysis and management Diffusion of innovations Interlocking directorates and elite interlocks… Well-known ‘networkers’: Mark Granovetter (Stanford University), who argued and empirically demonstrated that economic action is necessarily embedded in the world of social relations and connections David Knoke (University of Minnesota). Together with James H.Kuklinski, he published the work on SNA cited above, “Network Analysis” (1982) Barry Wellman (University of Toronto). The author of one of the basic works on SNA statistical procedures – “Network Analysis: some basic principles” (1983) Stanley Wasserman – професор психології, статистики та соціології. Разом із K.Faust написав фундаментальну книгу з методології соціальних мереж “Social Network Analysis” (1994) … and plenty of others (mostly from the USA and the EU) Publications on SNA theory, methodology and empirical applications can be found at ... INSNA (International Network of Social Network Analysis). Founded 1978. Official web-page is www.insna.org Connections and Social Networks journals Summer schools in SNA: - Essex University Summer School in Data Analysis and Collection (UK) - ECPR Summer School in methods and techniques (Ljubljuana , Slovenia) Software UCINET & NetDraw is developed by Steve Borgatti (Boston college) and Martin Everett (University of Manchester) www.insna.org , www.analytictech.com – 1 month free-of-charge trial version is available Pajek (for Large Network Analysis) is developed by Vladimir Batagelj and Andrej Mrvar(Ljubljana, Slovenia) http://pajek.imfm.si/doku.php?id=download – free-of-charge InFlow – mostly used for the organizational network analysis is developed by Valdis Krebs (Cleveland, Ohio) www.orgnet.com