NUK - logo
FSS, J. Goričar Central Social Sciences Library, Ljubljana (ODKLJ)
Opening hours: Mondays to Fridays 8 a.m. - 8 p.m., Saturdays 8 a.m. - 2 p.m..
 More... 

Subject Headings    Textbook    Zebra   Research guides
  • Generalized blockmodeling
    Ferligoj, Anuška ; Doreian, Patrick ; Batagelj, Vladimir
    Network analysis has attracted considerable interest from social and behavioral science community in recent decades. Much of this interest can be attributed to the focus of network analysis on ... relationships among units, and on the patterns and implications of these relationships. Social networks consist of social actors with one or more social relations defined over the actors. Social actors can be, e.g., individuals, groups, organizations. In social network analysis attention is focused on relation which is a set of the relationships (ties) between social actors. A relation among workers in an organization can be "to communicate with". The relation can be symmetric (e.g., coauthorship) or non-symmetric (e.g., friendship). A social network canbe presented by a set of related pairs of units, or a relational matrix, ora graph. We distinguish between one-mode and two-mode networks. One-mode social network is defined by the set of social actors and the relationships defined only between them, e.g., friendship relation among pupils in a class. If there are two sets of units and the relationships are defined between units of the first set and units of the second set we call it a two-mode social network. An example is membership network, e.g., scientists (first set) are members of different scientific associations (second set). We can collect data on whole or egocentered networks. When the ties for each pair of units are known we have a whole network. If a set of actors are given (e.g., a random sample) and only ties from each of these actors (egos) to some actors (alters) are measured (usually not also ties between these alters) we speak about egocentered networks or personal networks. There are several ways how to obtain social network data, e.g., survey and questionnaire methods, archival network data collection, observation, and experiments. The most frequent data collection method is survey method where several network specific methodological problems exist, e.g., the specification of the network boundaries, name generator instruments for egocentric networks (recognition vs. free recall, free vs. fixed choices),interview context effects, interviewer effects, name interpreters, data collection mode, networkdata quality (accuracy, reliability, validity). Today most networks areproduced from data already stored in data bases or available on the internet. There are several methods and approaches to social network analysis when analyzing whole networks: clustering (e.g., component, clique, k-core, cut, island), centrality: actor centrality (degree centrality, closeness centrality, and betweeness centrality) and group centralization, blockmodeling, signed networks, statistical models that represent the dependence between ties (for cross-sectional data: exponential random graph p*models and for longitudinal data: actor-oriented models and tie-oriented models) and other approaches. Recent developments of models and methods is social network analysis can be found in Carrington, Scott, and Wasserman (2005). One of the major goals of social network analysis is to discern fundamental structures of networks in ways that allow us to get insight into the structure of a network and to facilitate our understanding of network phenomena. The most used tools for doing this is blockmodeling, a collection of methods for partitioning networks according to well-specified criteria. Thegoal of blockmodeling is to reduce a large, potentially incoherent network to a smaller comprehensible structure that can be interpreted more readily. Blockmodeling, as an empirical procedure, is based on the idea that units in anetwork can be grouped according to the extent to which they are equivalent, under some meaningful definition of equivalence. In the talk the optimizational approach to blockmodeling proposed by Doreian, Batagelj, and Ferligoj (2005) is discussed. Methods where a set of observed relations are fitted to a pre-specified blockmodel are presented and several examples are given. At the end some further extensions are discussed (e.g., Batagelj, Ferligoj, and Doreian 2007). The described (generalized) blockmodeling methods are implemented in program Pajek (Batagelj, Mrvar, Ferligoj, and Doreian 2004; de Nooy, Mrvar, and Batagelj 2005).
    Source: Workshop on blockmodeling : program and abstracts (Str. 5-6)
    Type of material - conference contribution
    Publish date - 2008
    Language - slovenian
    COBISS.SI-ID - 26998621

Shelf entry