Exponenciální modely náhodných grafů : modelování relačních mechanismů na případu sítě organizací zapojených v českém uhelném sektoru
| Title in English | Exponential Random Graph Models : Modelling Relational Mechanisms in the Inter-organisational Network of the Czech Coal Subsystem |
|---|---|
| Authors | |
| Year of publication | 2024 |
| Type | Article in Periodical |
| Magazine / Source | Sociologický časopis |
| MU Faculty or unit | |
| Citation | |
| Web | https://sreview.soc.cas.cz/corproof.php?tartkey=csr-000000-0292 |
| Doi | http://dx.doi.org/10.13060/csr.2023.046 |
| Keywords | social network analysis; exponential random graph models; political networks; social mechanisms; statistical models |
| Attached files | |
| Description | This study provides the first comprehensive introduction to exponential random graph models (ERGM) in the Czech academic literature. In it we apply ERGM to a network of 68 organisations involved in the Czech coal policy subsystem. First, we summarise the major limitations of the statistical modelling of network data arising from the interdependencies among observations and explain principled solutions to them provided by ERGM. Next, we discuss ERGM’s metatheoretical assumptions and their embeddedness within the broader context of social science research. We then introduce three types of relational mechanisms (endogenous, individual, and dyadic) operationalised as specific configurations, which we illustrate through the empirical example of an expert information network. Following a descriptive analysis we apply ERGM, breaking it down into three main steps: simulation, estimation, and estimation assessment. We provide a detailed interpretation of the model’s development and results, along with recommendations for building a model and solutions to convergence failure problems. One important finding is that one predictor of the exchange of expert information is ideological homophily, which reduces the potential of expertise to seek compromise solutions. We close with a discussion of the results and ERGM extensions to apply to more complex types of network data such as bipartite and multiplex networks and valued and longitudinal data. |
| Related projects: |