Social Networks Agentbased Modelling and Social Network Analysis
Social Networks: Agent-based Modelling and Social Network Analysis with PAJEK Richard Taylor¹ and Gindo Tampubolon² ¹Centre for Policy Modelling, Manchester Metropolitan University ²Centre for Research on Innovation and Competition, University of Manchester ESRC Research Methods Festival, Oxford, 17 th-20 th July 2006, & Oxford Spring School, Dept. of Politics and International Relations
Methodology of ABM Objectives: • Characterisation of possible outcomes rather than prediction • Understanding of processes that lead to outcomes Steps • Problem formulation: observations, hypotheses, theory • Abstraction to create conceptual model (expressed as equations, set theory notation, or precise description) • ABM implementation • Design of experiments (parameter choices, batch sizes) … continued
Methodology of ABM (2) Steps • Analysis of results. Verification of implementation • Comparison of simulation outcomes with observations and/or with analytical solutions • Validation step - cross validation on both the microbehaviours and the macro outcomes / signatures (steps are iterated) Additional steps: • Model docking and different grains of analysis • Dissemination in sufficient reproducible detail
When using real networks as inputs to an ABM Do network structural dynamics seem important or not? • Are there any longitudinal data on how networks evolve? • Are there any findings/descriptions of how they evolve?
When using simulationgenerated networks • Are they emerging from some sociologically-founded process? • Use with care random / regular / small-world / scale-free graphs which although convenient may not be appropriate models • What are the appropriate measures by which generated networks could be analysed?
When using networks and grids Spatial grids are commonly used because for many types of interaction, spatial location is important • Grid models are relatively easy to set up • Network models can also be placed on a grid Consider, • Is there sufficient justification for doing so? • What additional complexity is thus generated?
Discussion point Typically we ask: What network types emerge from different social processes? Based on generative ABM, similarly we might ask: How do existing networks constrain social processes?
PAJEK vs. UCINET Similarities and differences between PAJEK and UCINET: • UCINET is free-to-try whereas PAJEK is free • UCINET is an older project, more widely used and better integrated with other software Both software supposed to import and export both file types (but I have found this doesn’t work) DRAW / NETDRAW are plug-ins for these packages that do network visualisations via several layout algorithms
- Slides: 8