D02-01PM: Inferential Network Analysis

Inferential Network Analysis introduces statistical methods for analyzing networks, which are sets of nodes and their connecting ties. Networks model phenomena like policy networks, lobbying coalitions, governance systems, international relations, financial deals, and policy diffusion. These systems evolve in complex ways, and researchers aim to understand tie formation and behavior adoption within networks. Key questions include: 1) How to model datasets with non-independent observations? 2) How to explain connections or node characteristics using covariates and theories of data endogeneity? 3) How to simulate processes to predict network evolution?
The course introduces statistical models for explaining and predicting tie formation using node characteristics, ties, and the network. It covers the exponential random graph model (ERGM), its specification, estimation, and implementation in R, as well as extensions for temporal and valued relations. Alternatives like latent space models (LSM), quadratic assignment procedure (QAP), stochastic actor-oriented model (SAOM), latent-order logistic (LOLOG) model, relational event model (REM), and network autocorrelation models (NAM) are also discussed. All models are covered theoretically and practically using R, with daily assignments. Participants will confidently choose, implement, and interpret models, assess fit, simulate networks, and diagnose issues. Familiarity with logistic regression and R is recommended. Participants should ideally bring laptops.

Contents:

The course Inferential Network Analysis introduces statistical methods for analysing networks. Networks, or graphs, are sets of nodes and their connecting ties. Networks model complex phenomena, such as policy networks among political actors; lobbying ties of interest groups to policymakers or lobbying coalitions; recruitment of experts into international organizations; patronage relationships; multilevel or collaborative governance systems; international relations, including conflict, alliances, trade, and migration; financial deals; political debates; or the diffusion of policies among states. Networks are also prevalent in sociology, economics, ecology, biology, physics, public health, and many other fields, and the course is open to participants from any field as well as data science practitioners.

Networks evolve in complex ways, and researchers aim to understand the formation of ties at the micro level to explain system evolution and test theories of network formation. Researchers also study behaviour adoption as a consequence of being embedded in a network. Central questions are: 1) How can we model any datasets where the observations are not independent and identically distributed (i.i.d.)? 2) How can we explain and model connections between nodes (or characteristics of nodes) using covariate data and theories about the endogeneity in the data? And 3) how can we simulate such processes forward in time to predict future states of the network or the characteristics of the nodes?

The course introduces statistical models for explaining and predicting tie formation using node characteristics, ties, and networks. It covers the exponential random graph model (ERGM) as the workhorse of statistical network analysis, including specification, estimation, and implementation in R. Extensions for temporal and valued relations and alternative models are also discussed, such as the Temporal ERGM (TERGM), latent space models (LSM) including Additive and Multiplicative Effects for Networks (AMEN), the quadratic assignment procedure (QAP), the stochastic actor-oriented model (SAOM), latent-order logistic (LOLOG) models, relational event models (REM), and network autocorrelation models (NAM).

All models will be discussed theoretically, in application, and practically using R. Participants will solve daily assignments after class to maximise learning. A working knowledge of logistic regression and familiarity with R are recommended. Participants should ideally bring laptops.

Learning outcomes:

Participants will confidently choose, implement, and interpret models, assess fit, simulate networks, and diagnose issues.

Admission requirements:

Participants should be interested in thinking about social phenomena as complex networks. Familiarity with R is recommended for the software parts, though the theory sections will still be engaging. Participants should ideally bring their own laptops.

Daily schedule:

Monday:

• Introduction and why we need network-specific models
• Network dependence
• Working with network data
• Software lab: Networks in R

Tuesday:

• Introduction to the ERGM
• Form, specification, and interpretation of ERGMs
• Goodness-of-fit assessment for ERGMs
• Model degeneracy
• Estimation, sampling, and simulation
• Software lab: ERGMs in R

Wednesday:

• Empirical ERGM examples
• Quadratic assignment procedure (QAP)
• Temporal ERGM (TERGM)
• Software lab: ERGMs in R; TERGM; QAP

Thursday:

• LSM and AMEN
• SAOM (if time permits)
• LOLOG (if time permits)
• Software lab: LSM and LOLOG in R

Friday:

• NAM
• REM
• Student presentations (voluntary)
• Software lab: REM in R

This description is subject to change at the discretion of the Instructor

2 Credits

For completion of all work before and during the course, as outlined by the Instructor, and 90% participation and attendance of the course.

2 Additional Credits

Course specific extra assignments. These can include submitting assignments before the course, daily assignments, and/or a final assignment to be completed after the course as decided by the Instructor.