Methodological developments in the field of network analysis

Bibliometric network analysis is based on the application of mathematical and statistical methods in general and network analysis methods in particular within the framework of scientometric research. This methodology involves building networks that connect different bibliometric entities (publications, authors, journals, keywords, etc.) with relationships of various types (co-presence, citation, co-citation, bibliographic combination, etc.). The sources of data for bibliometric analysis are publications in the form of bibliographic descriptions presented in various bibliographic databases.

The expertise of the laboratory team concerns working with Web of Science, Scopus, OpenAlex, and eLibrary databases as sources of information. The methodological and instrumental solutions presented by the laboratory team make it possible to convert large data sets of bibliographic descriptions into files for working in programs for network analysis, for example, Pajek (WoS2Pajek, OpenAlex2Pajek, Bib-eLib programs). Bibliometric network analysis is also implemented using the VOSviewer and Biblioshiny tools widely used in the scientific community. Laboratory members are engaged in comparing various bibliographic databases, as well as methodological approaches and tools for working with them (for example, comparing WoS and eLibrary data using the example of a dataset on Russian sociologists; comparing the functionality of the Pajek and VOSviewer programs). Methodological developments are also associated with the construction of various types of bibliographic networks, incl. changing over time. The laboratory team sets itself an ambitious goal to develop and implement a methodology for scientometric analysis of Russian-language scientific publications.

The methodology of bibliometric network analysis is actively used by laboratory team to study the development of various scientific fields and disciplines, their thematic structure and scientific teams, which is reflected in a large number of publications in Russian and English.

Blockmodeling methodology, developed specifically for clustering network data, allows you to cluster nodes that have similar patterns of relationships with other nodes and interpret the structure of relationships between the resulting clusters.

The scientific advisors of the laboratory, Anuska Ferligoj and Vladimir Batagelj, known as one of the founders of blockmodeling, are engaged in various new developments in this direction, presenting new methods and algorithms for blockmodeling for various types of network data and actively applying them in empirical research. In Russian practice, the blockmodeling method was described and applied to study the structure of the scientific community of sociologists by the Laboratory team.

Publications:

• Doreian P, Batagelj V, Ferligoj A. Generalized Blockmodeling. Cambridge University Press; 2004.
• Doreian, Patrick, Vladimir Batagelj, and Anuska Ferligoj, eds. Advances in network clustering and blockmodeling. John Wiley & Sons, 2020.
• Shcheglova T. E., Maltseva D. V., Kim A. V. Blockmodeling for the analysis of social structures: methodological foundations // Sociology: methodology, methods, mathematical modeling. 2021. No. 52. P. 7-35. doi
• Kim A. V., Maltseva D. V., Shcheglova T. E. Blockmodeling for the analysis of social structures: an example of studying the structure of a community of St. Petersburg sociologists // Sociology: methodology, methods, mathematical modeling. 2021. No. 53. P. 7-38. doi

In the context of the development of network analysis, the study of networks in dynamics is becoming increasingly important for understanding complex relationships. Recognizing this need, researchers have resorted to developing new methodologies for analyzing and constructing networks — particularly in the field of network modeling.

Exponential Random Graph Modeling (ERGM) allows us to more deeply explore various phenomena that arise from the complex interactions and behavior of network participants. Such models are used to explain the global structure of a network while allowing inferences to be made about the behavior of links at the micro level. By comparing a network with a large number of random networks formed on the basis of topological characteristics or individual attributes, we can recognize those local patterns that underlie global processes. ERGMs allow you to study and statistically evaluate various network phenomena – clustering, homophily, autocorrelation. Using these types of models, we can figure out how and why scientists collaborate with each other and understand what makes them work together. Temporal exponential random graph models (TERGMs) add a temporal dimension to the study of these processes.

Actor-oriented stochastic models (Stochastic Actor-Oriented Models, SAOMs) represent one of the most developing and promising means of analyzing the mechanisms of social development, relationships and evolution of various networks. SAOM acts as a lens through which researchers can decipher network dynamics, revealing the underlying processes driving network evolution. By bridging the gap between the observable and the unobservable, the model provides a valuable tool for network analysts seeking to uncover the hidden patterns and mechanisms that govern social interactions and network structures.

Network modeling is used in research projects of laboratory members to identify hidden patterns of interactions in educational groups, as well as communities of scientists.

The methodology for analyzing social networks involves the study of various types of networks: complete or ego networks, with directed and undirected connections, dynamic networks that change over time (temporal), spatial networks, networks with different weights of connections, several types of connections (multi-relational) and nodes (2-mode networks), multilevel networks and other specialized networks. Various types of networks are used in research projects of laboratory team, depending on the research goals and objectives. At the same time, our team members are developing new types of networks, such as temporal networks for bibliometric network analysis, multiway networks (Vladimir Batagelj). Recent development by Prof. Vladimir Batagelj concerns the use of Sequence analysis methods for network analysis (using the example of biographies of politicians).

Ego-network analysis is a technique used in social network analysis to study the relationships surrounding a central personality (ego) within a network. By focusing on the ego and its immediate connections (alters), one can gain insight into the structure, dynamics, and patterns of influence in small social circles.

This approach allows for a deeper understanding of how people interact, exchange information, and influence each other in their immediate social environment. Ego network analysis often involves examining various attributes of connections, such as strength and directionality, to reveal features of interpersonal relationships. Insights from ego network analysis can reveal features of social interactions at both the micro and macro levels. Ego network analysis can be of interest to research in a variety of disciplines, including sociology, psychology, anthropology, and organizational behavior.