Dynamic Network Actor Models (DyNAMs): An actor-oriented framework for studying time-stamped network data
Christoph Stadtfeld | Switzerland
Ample theoretical work on social networks is explicitly or implicitly concerned with the role of interpersonal interaction. However, empirical studies to date mostly focus on the analysis of stable relations. This talk introduces Dynamic Network Actor Models (DyNAMs) for the study of interpersonal interaction through time.
The presented model addresses three important aspects of interpersonal interaction. First, interactions unfold in a larger social context and depend on complex structures in social systems. Second, interactions emanate from individuals and are based on personal preferences, restricted by the available interaction opportunities. Third, sequences of interactions develop dynamically, where timing of interactions relative to one another contains useful information.
It is further discussed how DyNAMs can be applied to study coordination networks through time. It is taken into account that the creation of coordination ties between two actors is typically the outcome a two-sided agreement process in which both actors’ opportunities and preferences are aligned.
The DyNAM framework is conceptually and empirically compared to the relational event model, a widely used statistical method for the study of social interaction data.
Stadtfeld, C.; Hollway, J. & Block, P. Dynamic Network Actor Models: Investigating Coordination Ties through Time. Sociological Methodology, 2017, 47 (forthcoming)
Stadtfeld, C. & Block, P. Interactions, Actors and Time: Dynamic Network Actor Models for Relational Events. Sociological Science, 2017 (forthcoming)
Butts, C. T. A Relational Event Framework for Social Action. Sociological Methodology, 2008, 38, 155-200
Some days are better than others: Examining time-specific variation in the structuring of interorganizational networks
Viviana Amati | University of Konstanz | Germany
Using longitudinal data we have collected on a set of more than 8,000 relational events connecting the members of a small community of health care organizations, we explore patterns of time variation in the effect of network mechanisms on the dynamics of interorganizational relations. Data are analysed by using an event-oriented model based on the assumption that the observed sequence of relational events is the outcome of a marked temporal point process. The analysis supports conclusions that are generally consistent with prior research showing that interorganziational relations are patterned consistently and systematically by tendencies toward reciprocity, assortativity, closure, and by inertial forces that tend to stabilize interorganizational collaboration. However, the analysis also reveals that the effects of local network mechanisms on relational events connecting organizations display significant time variation, and tend to operate differently at different points in time. We discuss the implications of this finding for our theoretical understanding of interorganizational networks as emergent from the interaction of time-invariant relational mechanisms and time-specific local contingencies.
Predicting Relational Events
Laurence Brandenberger | University of Bern & Eawag | Switzerland
Models for dynamic network analysis are becoming increasingly popular. Among such temporal network models are relational event models, where sequences of relational events are examined across time. Each of these events represents an edge (or tie) forming in a network at a distinct point in time. This flexible and dynamic form of network inference can be used to examine how actors behave in changing network settings. Examples of event networks include states co-signing agreements, parliamentarians bargaining over new regulations, or individuals interacting online. The additional information regarding the timing of events allow for a more precise estimation of popular network effects, such as popularity, triadic closure or homophily effects. Inference on how networks evolve over time can be gained from combining network effects with statistical models from survival analysis, such as conditional logistic regressions or Cox models.
However, estimated parameters may suffer from a form of omitted variable bias if the temporal dependencies are not specified correctly and/or sufficiently, resulting in a misspecification of the joint likelihood of the model.
This paper presents a simple approach to predicting relational events as well as goodness-of-fit measures to evaluate the simulated sequences and to determine which temporal dependencies are crucial to the data-generating process of event sequences.
Collaboration between Software Developers and the Impact of Proximity
Dawn Foster | University of Greenwich | United Kingdom
This study investigates collaboration in an open source software community using proximity theory as the theoretical lens with social network analysis and modeling of activities over time to predict collaboration.
Actors in this study are part of the Linux kernel community where they collaborate on one or more sub-projects using mailing lists as the primary method of collaboration. Collaboration occurs in real-time between actors that contribute to multiple sub-projects, work for firms that pay them to contribute to the Linux kernel, and are working virtually from locations across the globe. This complex setting can be better understood by using several dimensions of proximity: organizational, cognitive, institutional, social, and geographical. Collaboration is analysed using data from source code contributions and mailing list participation.
Open source software is developed in the open where anyone can view the source code and anyone with the knowledge to do so can contribute to the project. With no central group responsible for coordination of tasks, collaboration on the development of this software is emergent. Because people from around the world work on these projects together using online tools with publicly accessible interactions between people, it is a relevant setting for using social network analysis to understand and model network relationships.