Marlyne Meijerink - Roger Leenders - Joris Mulder

Learning gradual changes of social interaction behavior in temporal networks

This PhD project focuses on developing methods for modeling real-life social interaction dynamics. That includes the introduction of relational event models for studying social interaction dynamics to psychology researchers, investigating if and how drivers of social interaction develop over time, and studying how the duration of past interactions both impacts the future interaction rate and is impacted by past interaction patterns.

Discovering trends of social interaction over time: An introduction to relational event modeling.

The relational event framework provides a flexible approach to studying the mechanisms that drive how a sequence of social interactions evolves over time. This paper presents an introduction of this new statistical framework and two of its extensions for psychological researchers. The relational event framework is illustrated with an exemplary study on social interactions between freshmen students at the start of their new studies. We show how the framework can be used to study (a) which predictors are important drivers of social interactions between freshmen students who start interacting at zero acquaintance, (b) how the effects of predictors change over time as acquaintance increases, and (c) the dynamics between the different settings in which students interact.

Dynamic relational event modeling: Testing, exploring, and applying.

The relational event model (REM) facilitates the study of network evolution in relational event history data, i.e., time-ordered sequences of social interactions. In real-life social networks, however, it is likely that network effects, i.e., the parameters that quantify the relative importance of drivers of these social interaction sequences, change over time. In these networks, the basic REM is not appropriate to understand what drives network evolution. Therefore, this research extends the REM framework with approaches for testing and exploring time-varying network effects. Firstly, we develop a Bayesian approach to test whether network effects change during the study period or not. A simulation study was conducted that illustrates that the Bayesian test accurately quantifies the evidence between a basic (‘static’) REM or a dynamic REM. Secondly, in the case of the latter, time-varying network effects can be studied by means of a moving window that slides over the relational event history. A simulation study was conducted that illustrates that the accuracy and precision of the estimates depend on the window width: narrower windows result in greater accuracy at the cost of lower precision. Thirdly, one challenge of the moving window REM is to determine the window width that can best capture the time-varying network effects. Therefore, we develop a Bayesian approach for determining window widths using the empirical network data. A simulation study was conducted that illustrates that estimation with empirically determined window widths achieves both good accuracy for time intervals with important changes and good precision for time intervals with hardly any changes in the effects. Finally, in an empirical application, it is demonstrated how the approaches in this research can be used to test for and explore time-varying network effects of face-to-face contacts at the workplace.

Modeling the frequency and duration of social interaction (working title).

Relational event modeling approaches enable researchers to draw conclusions on the drivers of real-life social interactions (“events”). The duration of an event may contain important information on real-life social interaction patterns. For example, the duration of social contacts is considered important in understanding how information and disease spread among a group of individuals. Therefore, this research extends the relational event modeling framework with approaches for accounting for and predicting event duration. First, we propose an approach that enables researchers to include the impact of event duration on future event occurrence. In an empirical application, it is demonstrated that accounting for the duration of events allows us to better predict future event occurrence. Second, we propose a method for modeling the predictors of event duration, given that the event has started. An empirical analysis illustrates important insights that can be obtained from predicting event duration.

Software package

REMSTATS

The R software package remstats was developed for efficient computation of statistics for analyzing relational event history data with relational event models (REMs). The software package includes a wide range of endogenous and exogenous statistics for a tie-oriented REM (Butts, 2008) or actor-oriented REM (DyNAM, Stadtfeld & Block, 2017). It can handle directed and undirected events and allows for REMs that consider event types in the dependent variable. The resulting object can be used with the R software packages remstimate or relevent (Butts, 2008) to obtain estimates for the model parameters. The package can be retrieved from GitHub.

Applications

Discovering trends of social interaction over time: An introduction to relational event modeling. Manuscript submitted for publication.

The CONNECT study is an extensive research in the joint development of personality and social relationships among university students. In our analyses of the CONNECT data, we investigate how interaction patterns develop as freshmen students get acquainted with each other and how personality affects the development of these interaction patterns. Results show that routinization (the tendency for students to keep interacting with the same partners) and transitivity (the tendency for students to interact with whom they have many past interaction partners in common) develop early, as the students are just acquainted, and remain relatively stable thereafter. The same effects were also found within a setting for social interaction: Students were more likely to interact again in a leisure or study-related setting with interaction partners with whom they have interacted more before in the same setting and were more likely to interact with others in a specific setting for social interaction when they have more past interaction partners in common in that same setting. Furthermore, it was shown that more extraverted students were attractive interaction partners, especially during the weekends. More agreeable students, however, were less often involved in interactions.

Dynamic relational event modeling: Testing, exploring, and applying.

In the empirical application we investigate how face-to-face interactions between employees of an organization develop over time. In particular, we were interested in how the effects of past interaction patterns on future interaction rates develops over time and how these effects impact the interaction rates when employees work in the same department versus in different departments. Results showed that employees were most likely to interact with others who worked in the same department, with whom they have interacted more recently, with others who worked in different departments and with whom they have interacted more in the past, and with those persons who interacted more with common others in the past. The strength of these effects varied within working days. The code for the analyses is available on request and will be available online soon.

Papers

Meijerink-Bosman, M., Leenders, R., & Mulder, J. (2022) Modeling the speed and duration of social interaction.

Meijerink-Bosman, M., Leenders, R., & Mulder, J. (2021) Dynamic relational event modeling: Testing, exploring, and applying.

Meijerink-Bosman, M., Back, M., Geukes, K., Leenders, R., & Mulder, J. (2021) Discovering trends of social interaction over time: An introduction to relational event modeling.

BFpack: Flexible Bayes Factor Testing of Scientific Theories in R. Journal of Statistical Software, 100(18), 1–63.