Public Lecture: Stochastic Network Modeling - Enhancing Interpretation and Facilitating Comparisons
A joint public lecture by the University of Groningen and the University of Hamburg
Foto: UHH/Esfandiari
Monday, 26 October 2026
10:00-12:00
University of Hamburg
The lecture is concerned with how social networks and related relational processes can be modeled and analyzed statistically. It focuses particularly on generative stochastic network models, introducing approaches such as Exponential Random Graph Models (ERGMs) and Stochastic Actor-Oriented Models (SAOMs), and discussing how effects in these models can be interpreted and compared across datasets and model specifications.
The first part will provide a general introduction to stochastic network modeling. The second part will focus on challenges in interpreting and comparing network effects, including recent developments involving probability-scale marginal effects and meta-analytic approaches. The presentations will be illustrated with empirical examples, including friendship networks among pupils and organizational networks.
The event is aimed at researchers, students, and practitioners with an interest in social network analysis and the stochastic modeling and statistical analysis of social networks and related relational processes.
Programme
Foundations of Stochastic Network Modeling
Christian Steglich, University of Groningen, will give a general introduction to stochastic network models such as ERGMs and SAOMs. These models bridge the gap between familiar (binary and multinomial) logistic regression and highly interdependent social network data. They make it possible to disentangle processes that are otherwise hard to separate: purely endogenous tendencies such as reciprocity and transitive closure (“friends of friends become friends”) and attribute-based effects such as gender or class homophily in friendship networks. Beyond personal networks, they are also used to understand the dynamics of interorganizational networks, for example in organizational fields.
Interpreting and Comparing Network Effects
Daniel Gotthardt, University of Hamburg, and Marijtje van Duijn, University of Groningen, will discuss how effects in stochastic network models can be interpreted and compared across datasets and model specifications. Because network effects are estimated on a latent scale, direct comparisons can be highly problematic, and the estimates are often hard to relate to the descriptive network measures researchers typically use. At the same time, there is a great interest in generalizing and comparing network processes, for example across school classes or organizational fields, and in understanding confounding and mediation between direct homophily effects and indirect drivers of segregation and polarization.
This part presents recent probability-scale marginal-effects-based approaches that address these interpretational difficulties and discusses the extent to which they yield more intuitive, comparable interpretations of network processes. It also discusses how these approaches could be integrated into a broader meta-analytic framework.
About the lecturers
Christian Steglich is Associate Professor of Sociology at the University of Groningen and the Institute for Analytical Sociology, Linköping University. His research focuses on statistical inference for social networks, social influence, and modeling macro- and meso-level dynamics. Since 2025, he acts as maintainer of the RSiena software package for the analysis of longitudinal social network data.
Marijtje van Duijn is Professor of Statistics at the Department of Sociology, University of Groningen. Her research focuses on the development and application of statistical models, with an emphasis on random effects (multilevel) models for non-normal data, with applications in social network analysis.
Daniel Gotthardt is Research Associate at the Professorship for Sociology, with a focus on Digital Social Science (Achim Oberg), Faculty of Business, Economics and Social Sciences, University of Hamburg. His research focuses on the formalization and statistical analysis of social processes, in particular relational processes in networks and groups. In his PhD, he models the dynamics of organizational fields amid a changing climate.
Practical information
Welcome reception: from 09:30
Get-together: 12:00-13:00, informal, for those who would like to join
Location: University of Hamburg; room to be announced
Participation: Free of charge
Registration: Please indicate your interest by email to daniel.gotthardt"AT"uni-hamburg.de.
Registration helps us plan the room capacity and notify interested participants when the exact location has been confirmed. Spontaneous attendance is also welcome, subject to available capacity.
The room and any further practical information will be added to this page. Please use this page as the current source of information about the event.
Organizers and Project
This event is organized jointly by:
- the Professorship for Sociology, with a focus on Digital Social Science (Achim Oberg), Faculty of Business, Economics and Social Sciences, University of Hamburg, and
- the Statistical Methods group in the Department of Sociology at the University of Groningen.
The Digital Social Science Team in Hamburg studies how social structures, processes, and dynamics emerge in digital data. As a team of organizational sociologists and business informaticians, it combines methods for capturing and analyzing large-scale web data with approaches from social and semantic network analysis, applying them to contemporary social dynamics such as digitalization, climate change, and developments in the field of science and higher education. Rather than applying these methods off the shelf, the team actively extends and adapts them to meet the theoretical and empirical requirements of analyzing organizational fields and their dynamics.
The Statistical Methods group at the University of Groningen has broad expertise in statistics and methodology and their application in empirical sociology. Its research focuses on models for social network analysis, particularly the development of stochastic actor-oriented models for the analysis of longitudinal social network data. These models are used to analyze school classrooms and organizational networks, and as simulation models to study, for instance, the development of the structure of one or more social networks and social influence processes.
This event is part of COMMODE (Comparing Effects in Stochastic Network Models). Supported by the Groningen-Hamburg Fund, this joint initiative between the Department of Sociology, University of Groningen, and the Faculty of Business, Economics and Social Sciences, University of Hamburg, aims to develop new methods for interpreting and comparing effects in stochastic actor-oriented models. The project seeks to share these methodological advancements, fostering additional research cooperation between method developers and applied researchers in various domains.