Stochastic network models for analytical sociology
| Title: |
Stochastic network models for analytical sociology |
| DNr: |
LiU-2019-9 |
| Project Type: |
LiU Compute |
| Principal Investigator: |
Christian Steglich <christian.steglich@liu.se> |
| Affiliation: |
Linköpings universitet |
| Duration: |
2019-03-29 – 2020-04-01 |
| Classification: |
50401 |
| Keywords: |
|
Abstract
Social network data consist of a "complex" ensemble of social actors and the interdependencies between them, typically expressed as binary relational variables. With the help of stochastic network models (e.g., exponential random graph models or stochastic actor-oriented network evolution models) it is possible to explain these complex data structures as emergent consequences of less complex rules of social interaction, formulated at the level of the social actor, or at the level of the dyad (i.e. pair of social actors). Fitting such models to existing data sets requires techniques of simulation-based inference. For example, the position of a cloud of simulated networks is calibrated such that the observed data set is in the middle of that cloud in some multidimensional projection.
The calibration of these models to larger data sets can pose a computational challenge. In addition, analytical sociology aims to explain not only a single observed network, but exploit the simulation framework also for generating networks under plausible but counterfactual model parametrisations. This way, the effect of social interventions can be played through without having to actually implement them in the real world. Also this task can be computationally effortful.
In this first small project, my aim is to gain experience with running such network analyses and network simulation studies on the SNIC infrastructure. I will upload and work with a few data sets I have already available (mainly studies on network interventions to combat smoking among adolescents in school cohort networks). Depending on success of a research proposal, I may get access to larger URL share data sets of which I would want to import small part to the SNIC environment for network analysis.
