Mitchell Centre Seminar Series - Scott Duxbury (University of North Carolina, Chapel Hill) - General Procedures for Micro-Macro Network Analysis
| Dates: | 10 December 2025 |
|---|
| Times: | 15:00 - 16:30 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | School of Social Sciences |
|---|
| Speaker: | Scott Duxbury |
|---|
|
|
How do interpersonal social interactions generate novel network structures? Historically researchers examined micro-macro linkages in social networks using simulation approaches, especially empirically calibrated agent-based models grounded in generative models for network structure. The research presentation will review recent developments in statistical procedures for testing micro-macro linkages. The micro effect on macro structure (MEMS) approach defines target quantities that capture the total contributions of a micro selection mechanism on a novel macro structure. The talk will introduce the MEMS, describe data and modeling requirements, and estimation procedures. It will then dedicate considerable time to positioning the MEMS in the counterfactual framework for causal inference. It will report new results showing that the MEMS has close relationships to known causal quantities, specifically the attributable effect of Rosenbaum (2001) and the population average overall causal effect in the treatment program framework of Halloran and Hudgens (2008). It will also show that the MEMs can be decomposed into direct, total, and indirect effects under ignorability assumptions common to causal mediation analysis, and that the MEMS supports a causal interpretation under the "causes of effects" framework. Implications for conducting causal inference using the MEMS are discussed, along with limitations of the current framework and directions for further research.
Speaker
Scott Duxbury
Role: Associate Professor
Organisation: University of North Carolina, Chapel Hill
Travel and Contact Information
Find event
5.212
University Place
Manchester
