In-person workshops

  • Select title for full details, including co-presenters and workshop description.
    • Each workshop is limited to 30 participants, unless otherwise stated in the description.

 

    • Registration for in-person workshops will be available as you register for Sunbelt 2024.

 

  • You must register for the main conference in order to register for in-person workshops.
    •  
Monday, 24th June AM Half Day

Kariyushi Rao, Ty Hayes, Oana Vuculescu, Mads Kock Pedersen, Carsten Bergenholtz

 

Web application-based group experiments are a powerful tool to understand social networks. The controlled, yet dynamic, environments provided by web applications allow for the testing of complex theories about social structures, influence, and information flow in ways that were previously infeasible. This workshop aims to equip researchers with the necessary knowledge and skills to navigate the intricacies of designing and implementing such studies effectively. More specifically, our seminar provides a high-level overview of the key challenges and concerns related to the design, architecture, and execution of web application experimental studies with interactive groups. One of the key challenges researchers face is selecting the right platform for their experimental studies. We present three main options: (1) Out-of-the-box solutions that require minimal customization, (2) the DIY approach where researchers can build upon open-source platforms like oTree and Empirica, and (3) the fully custom route where researchers either develop the application themselves, or contract third-party developers. It is imperative for researchers to ask pertinent questions early in the design process as these platform decisions constrain what experiment designs can be realized. Thus, when we as researchers are faced with determining the most appropriate choice among these alternatives, the decision largely hinges on three criteria:
 
(1) Technological Complexity: Each option comes with its own set of limitations. For instance, out-of-the-box solutions might lack flexibility, whereas fully custom routes might present steep learning curves or dependency on third parties.
 
(2) Costs: While ready-made solutions might seem cost-effective initially, custom platforms, despite their higher upfront costs, might offer long-term value, especially for complex experiments.
 
(3) General Architecture: The underlying structure of the chosen platform should align with the experiment’s requirements. DIY options, for instance, allow for a good balance of customization and established architecture, making them suitable for many researchers.
 
We take a deeper dive into the most common issues researchers confront, and provide illustrative examples of solutions from our own work. At the design stage, we discuss the data collection and capture, cross-device compatibility, and developing effective pilot tests. On the subject of architecture, we discuss efficient participant grouping mechanisms, best practices for minimizing attrition, and how to deal with attrition when it does occur. Finally, we offer advice for successful execution, including pre-screening methods to protect against bad actors, live monitoring, and designing payment schemes for maximum retention and attention.
 
Workshop Length: 3 hours.
 
Maximum Attendees: +/– 100 (flexible).
 

Martina Morris

 

This workshop provides a hands-on tutorial to using exponential-family random graph models (ERGMs) for statistical analysis of social networks, using the “ergm” package in statnet. The ergm package provides tools for the specification, estimation, assessment and simulation of ERGMs that incorporate the complex dependencies within networks. Topics covered in this workshop include:
 
• an overview of the ERGM framework;
 
• types of terms used in ERGMs
 
• defining and fitting models to empirical data;
 
• interpreting model coefficients;
 
• goodness-of-fit and model adequacy checking;
 
• simulation of networks using fitted ERG models;
 
• degeneracy assessment and avoidance.
 
Duration: 3 hrs
 
Prerequisites:
 
Familiarity with R. If new to network analysis, the workshop Introduction to SNA with R and statnet is strongly suggested.
 
Website for further info: https://statnet.org/workshop-ergm/
 

Michał Bojanowski, Lorien Jasny

 

Those wishing to use the R programming language for network analysis now have a plethora of choices when it comes to libraries. In this workshop, we survey the main packages used for network data management, analysis, and visualization. We will cover 1) importing network data (from actual files), 2) network objects and attributes, 3) computing basic descriptives (attribute distribution, mixing matrix, density, degrees, betweenness, closeness), and 4) visualization (layouts, node aesthetics). These will be done side by side for the different packages, as well as discussion of the strengths and weaknesses of each. We conclude with time for attendees to work either on toy datasets or with their own data with help from instructors. This workshop is a unification of workshops “Using R and ‘igraph’ for Social Network Analysis” and “Introduction to Social Network Analysis with R and statnet” that has been offered on Sunbelt and EUSN conferences since 2011. It will serve as an introduction for those wishing to take “Moving beyond descriptives”, “Using ‘igraph’ for SNA: advanced topics”, “An introduction to ERGM with Statnet”, or other Statnet-related workshops on the program.

 
3hrs, at most 30 ppl
 

Scott Duxbury, Jenna Wertsching

 

Exponential random graph models (ERGM) are widely used in the social sciences to examine determinants of graph structure. This 3-hour workshop will introduce attendees to mediation and moderation analysis in ERGM using the ergMargins package for R. The workshop will describe why ERGM coefficients cannot be compared between models and why coefficients for interactions—including node matching, node mixing, and other common measures of homophily and heterophily—cannot be interpreted without adjustment. Topics covered will include (1) mediation analysis, (2) moderation analysis, (3) mediation analysis when the mediator is an interaction, and (4) mediation analysis when the main effect is an interaction. We will review a range of special cases, including interactions involving both continuous and discrete variables, necessary conditions for a causal interpretation, and mediation analysis involving endogenous graph statistics. Attendees will come away from the workshop with a deeper understanding of inferential difficulties in ERGM and with knowledge on how to address each issue using ergMargins.

 

Pablo Galaso, Sergio Palomeque

 

This workshop aims to provide practical training in the use of patent data for the study of collaborative networks. Patent data is a source of information on invention developments at the level of regions, countries, cities, or other types of sub-national dimensions. One of the difficulties of this type of administrative records is that patent offices do not assign a unique identifier to inventors or owners and, for this reason, the analysis of linkages between agents is not always possible. In recent years, various efforts have been made to disambiguate patents. Among them, the PatentsView project stands out (https://patentsview.org/), which works with patents registered at the United States Patent and Trademark Office (USPTO) by applying an algorithm that analyses each registration and determines whether two agents can be the same. From this information, it is possible to construct incidence matrices linking inventors, owners (i.e. firms and organisations), cities, regions, countries or even technologies. These data may reflect either co-partnership or co-authorship between actors (e.g. two inventors involved in the same patent). However, this does not necessarily imply collaboration, which is why some scholars propose to apply Back Bone Extraction (BBE) techniques that allows to identify significant links to approach the phenomenon of collaboration.
 
The workshop will provide participants with information on how to access the raw data available in the PatentsView platform, different ways of systematising this information to build networks, some BBE techniques to define meaningful links and certain specificities of the study of the topological structure in this type of networks. An overview of the meaning and usefulness of an important part of the tables available on the platform will be given. Finally, examples of research works that have used them will be presented and some of the general limitations of this type of data, as well as those particular to PatentsView, will be discussed. Special emphasis will be placed on the technologies associated with patents and how to use them from a social network analysis perspective, allowing this methodology to be applied to units of analysis that are not people or organisations, but rather technological fields.
 
The specific objectives offered by the workshop to its participants are: (i) to become familiar with patent data and its usefulness for the study of network analysis, (ii) to become familiar with the main literature on the subject, (iii) to download USPTO data available through the PatentsView platform, (iv) to learn about and process the different tables included in the database, (v) to build different types of collaborative networks from this data, and (vi) to analyse these networks, emphasising some stylised facts common to this type of interactions.
 

Monday, 24th June PM Half Day

Michał Bojanowski, Pavel N. Krivitsky

 

This workshop provides an introduction to analyzing egocentrically sampled data with exponential-family random graph models (ERGMs) for statistical network analysis. It is a hands-on workshop demonstrating how to fit, diagnose and simulate both static and dynamic ERG models from such data, using the ergm.ego package, part of the integrated Statnet software collection in R. Topics covered in this session include:

– a review of approaches to analyzing egocentrically sampled data,
– an overview of the statistical theory that supports the use of ERGMs for egocentrically sampled networks;
– defining and fitting ERGMs to egocentric data;
– interpreting model coefficients;
– checking goodness-of-fit and model adequacy; and
– simulating complete networks from the specified ERG models.
 
3hrs, at most 30 ppl
 

Brennan Antone, Kyosuke Tanaka, Noshir Contractor

 

Interactive teaching methods help learners internalize key concepts about social networks and apply them in their social or professional lives. Interactive activities can introduce “network thinking” and connect it to real-world decisions. This 3-hour workshop will introduce five different interactive teaching exercises, provide instructional materials, and discuss their use. These exercises are designed for use in classes, or as stand-alone workshops for organizational or professional development. The first exercise, 6-DOS (Six Degrees of Separation), helps participants consider their network awareness and strategies to raise their awareness. The second exercise, RAD (Relational Analytics Dashboard), is a teaching case on how network-enabled people analytics can shape organizational decision-making. Third, MDT (My Dream Team) is a platform for helping participants assemble teams and make new connections. The fourth exercise, PNR (Personal Network Report), guides participants through reflecting on their own networks in a professional or academic context. Fifth, the ERGM Builder exercise is a teaching tool designed to help participants understand and reason about statistical models of networks. We will provide a live demonstration of each teaching exercise, as well as distribute slides/material and instructional guides for running each exercise. Additionally, we will discuss the role that interactive exercises can play as part of a larger social network curriculum.

Length of the workshop: 3 hours
Maximum number of attendees: 40

Martina Morris

 

This workshop and tutorial provide a hands-on introduction to working with temporal network data in Statnet: from exploratory data analysis and visualization to statistical modeling and simulation with Temporal Exponential-Family Random Graph Models (TERGMs). TERGMs are a broad, flexible class of models for representing the structure and dynamics observed in temporal networks. They can be used for both estimation from and simulation of dynamic network data. The topics covered in this workshop include:
 
• A brief overview of exploratory data analysis with temporal network data (using the Statnet packages ‘tsna’ for descriptive statistics and ‘ndtv’ to create network movies),
 
• Different types of dynamic network data (network panel data, a single cross-sectional network with link duration information, and cross-sectional, egocentrically sampled network data)
 
• Model estimation tools for each type of data using the Statnet package tergm
 
• Model diagnostics in tergm, and
 
• Simulating dynamic networks from fitted models with tergm.
 
Duration: 3 hrs
 
Prerequisites:
 
Familiarity with R, and experience with the Statnet packages for static network analysis (ergm, network, sna). If new to network analysis with Statnet, the ERGM workshop is strongly recommended. Previous experience with the Statnet packages for descriptive analysis of temporal networks (tsna, networkDynamic and ndtv) is helpful but not required.
 
Website for further info: https://statnet.org/workshop-tergm/
 

James Hollway, Jael Tan

 

From infectious diseases to innovations, from memes to norms, networks regularly influence how outcomes are distributed. This workshop introduces the many tools available in the ‘manynet’ and ‘migraph’ packages in R for studying network diffusion or contagion. First, we look at the tools available for simulating various contagion processes, including simple and complex diffusion. We extend these models to a range of different compartment models, e.g. SEIRS, that can better represent more varied processes, and suggest how to evaluate the fit of these simulations with observed diffusions. Second, we look at tools for measuring, describing, or inferring aspects of these processes, from hazard rates to thresholds. We show how they can be used on observed diffusions too, so please bring data from any salient projects you are working on. Lastly, we will explore how, with the rest of the tools available in ‘manynet’ and ‘migraph’, we can identify points of intervention to accelerate or obstruct diffusion.

Because these procedures are based on ‘manynet’, they work with many different network formats and types, including ‘igraph’ and ‘network’ classes, as well as directed and multimodal data. manynet offers a single R package to assist researchers with making, manipulating, and mapping many types of networks and assist teachers with practical tutorials. Users can quickly and easily add, delete, reformat, transform, split and join, and visualise networks with sensible but flexible defaults in minimal code. Familiarity with R and RStudio is recommended.

Scott Duxbury

 

What network selection mechanisms generate unique network structures and topologies? What are the implications of those selection mechanisms for individuals nested in social networks? This 3-hour workshop will introduce attendees to recent statistical frameworks for micro-macro network analysis using the netmediate package for R. The workshop will describe (1) procedures for estimating, interpreting, and null hypothesis testing the effects of micro-level selection mechanisms on macro-level network structures, and (2) introduce recent statistical procedures for evaluating the indirect effects of micro-level network selection mechanisms on individual and group level outcomes (e.g., adolescent smoking behavior). Topics covered will include (1) micro-macro analysis when the interest is treating a specific network structure (e.g., segregation, transitivity) as the dependent variable and (2) identification of indirect network selection effects on individual and group outcomes. We will review necessary assumptions for each type of analysis and strategies for meeting those assumptions in observational network data. Attendees will come away from the workshop with a deeper understanding of statistical procedures for micro-macro network analysis and how these procedures can be used to address research questions that implicate indirect network selection effects using netmediate.

 

Monday, 24th June Full Day (with lunch break)

Raffaele Vacca

This workshop is an introduction to the R programming language and its tools to represent, manipulate, and analyze egocentric or personal network data. Topics include: introduction to ego-network research and data; introduction to data structures and network objects in R; visualizing ego-networks; calculating measures on ego-network composition and structure; converting ego-network measures to R functions; applying these functions to many ego-networks. The workshop emphasizes R tidyverse packages for data science, showing how they can be used to easily conduct common operations in ego-network analysis and scale them up to large collections of networks. We’ll cover specific packages for network analysis (igraph, network, egor), data management (dplyr) and programming (purrr). No previous familiarity with R is required; participants only need a laptop with R and RStudio installed. This workshop has been taught for the past several years at different international conferences, including INSNA’s Sunbelt and EUSN meetings. It draws on concepts and methods discussed in “Conducting personal network research: A practical guide” by Christopher McCarty, Miranda Lubbers, Raffaele Vacca and José Luis Molina (Guilford Press). More details on the workshop’s materials and instructor are here: raffaelevacca.com/egonet-r.

Robert Wilhelm Krause

 

Unfortunately, among the great data we collect there is often something missing. Respondents skip the network part of the survey, our software messes up, or parts the network refuse participation entirely. Even worse, standard missing data treatments are problematic when applied to network data, because they rely on the independence assumption. The problem here is that this does not only lead to biased imputations, but worse, the imputations ignore the dependences in our data that we set out to study with networks in the first place.
 
In the workshop, I will start with an introduction into simple missing data handling techniques, their pros and cons, and their theoretical implications. We will then dive into more complex model based treatments.
 
There will be ample time for discussions and participants are encouraged to talk about their specific (missing) data problems.
 
Always have in mind: You always make a decision about missing. Less important than the option you choose is, that you make an informed choice.
 
Full day workshop
Max 20 attendees

Ernst C. Wit, Federica Bianchi, Martina Boschi, Edoardo Filippi-Mazzola, Ruta Juozaitiene, Alessandro Lomi

 

Advances in information technology have increased the availability of time-stamped relational data such as those produced by email exchanges or interaction through social media. Whereas the associated information flows could be aggregated into cross-sectional panels, the temporal ordering of the events frequently contains information that requires new models for the analysis of continuous-time interactions, subject to both endogenous and exogenous influences. The introduction of the Relational Event Model (REM) has been a major development that has led to further methodological improvements stimulated by new questions that REMs made possible. In this short course, we introduce the REM, define its core properties, and discuss why and how it has been considered useful in empirical research. Then we will focus on how advances in relational event modelling are upgrading the method for new applications.
 
1. Introduction to REMs
 
If a process consists of a sequence of temporally ordered events involving a sender and a receiver, such as email communication or bank transactions, the REM can be used to identify drivers of this process. It is based on event history modelling, in particular the Cox model, which allows for convenient and efficient estimation.
 
2. Extending traditional network statistics in REMs
 
In this part we describe how REMs accommodate a wide array of network statistics, including (i) degree-based metrics and (ii) their intensity-based counterparts. Moreover, REMs allow the distinction between (iii) short- and long-term network dynamics, as well as the representation of (iv) time-weighted network dependencies among past events.
 
3. Mixed effect additive REMs
 
We show how to extend REM formulations with non-linear specifications of endogenous effects, as well as time-varying influence of covariates on the event rate. Furthermore, we show how the incorporation of random effects can uncover latent heterogeneity associated with individual actors or groups of them.
 
4. Reciprocity and triadic effects revisited
 
Many behavioural models that examine reciprocal or triadic relationships often overlook the element of time sensitivity. Instead, they assume that, once a reciprocal or triadic relationship is initiated, reciprocal or triadic motives remain stable over time. We introduce a framework that enables the estimation of the dynamic structure of these effects.
 
5. Global covariates
 
Covariates in traditional REMs are monadic (such as age or outdegree) or dyadic (such as age-difference or reciprocity) in nature. It was impossible to fit global covariates (such as weather or time-of-day), because they would cancel out in the partial likelihood. We present a simple method that allows fitting such covariates.
 
6. REMs with millions of events
 
Until recently, REM’s applicability was limited to small to medium-sized datasets due its computational complexity. Recent advancements address such computational challenges. In this workshop, we will explore novel strategies for fitting REMs efficiently, even in dynamic networks with millions of events.
 
The workshop will feature short explanatory sessions of approximately 20 minutes, followed by 40 minutes of hands-on computer practicals. Participants are encouraged to bring their own laptop with Rstudio pre-installed.
 
It is planned as a full-day (6 hour) workshop. The maximum number of participants is 40
 
Filip Agneessens, Tomáš Diviák
 
This 6-hour workshop provides an overview of network measures, as well as a short intro into data collection and data management with R. The focus is on complete networks, although some topics might also be useful for analyzing egonetwork data. We will be using the xUCINET package in R to calculate these measures.
 
The course outline is as follows:
– Introduction to social networks, different types of networks (including two-mode/affiliation networks and valued networks)
– Different types of datastorage: adjacency matrices, nodelists and edgelists, and incorporating attributes
– Basic visualization
– Centrality measures (closeness, betweenness, eigenvector) and nodal measures of position (diversity/homophily in egonetwork)
– Whole network structural measures (density, centralization, average geodesic distance, reciprocity, transitivity, homophily)
– Subgroups, such as cliques, as well as community detection
– Advanced visualization
– Datacollection via surveys
 
Maximum 70 participants.
6 hours workshop.
 

Jan Fuhse

 

Theory matters! It guides our attention in research, it gives us expectations for empirical analysis, and it allows us to interpret results as examples of wider significance. Traditionally, network research focuses more on methods than on theory, leading to laments about the lack of theory. Over the last 30-35 years, there have been important advances in this regard. Now we have a variety of theoretical approaches to networks particularly from sociology (rational choice, analytical sociology, relational sociology etc.) available, as well as a number of middle-range theoretical concepts (social capital, network mechanisms). However, often enough, researchers do not know which concepts and approaches work well with their research.
 
The workshop gives an introduction and reflection into the general perspective of social network analysis, it offers an overview of the currently most important concepts and theoretical approaches to social networks, and provides for a forum for participants to discuss their own empirical research in relation to theory. The focus of the workshop lies on theories that give answers to the questions: What are social networks? Why, and how, do they matter for social phenomena?
 
The following topics will be covered:
 
‒ the general perspective of network research in the social sciences, with its difference to other approaches;
 
‒ what is theory, and how does it matter in the research process?
 
‒ action theory and social capital;
 
‒ varieties of relational sociology (inspired by pragmatism, symbolic interactionism, and by Harrison White);
 
‒ network mechanisms (foci-of-activity, homophily, institutionalized role patterns, reciprocity, transitivity, preferential attachment, social control, brokerage, access to information), the epistemological status of network mechanisms;
 
‒ methodology: which theoretical approach work with which methods?
 
‒ what concepts and theoretical approaches fit the attendants’ empirical research projects?
 
Much of the workshop will be run as presentations by the lecturer, complemented by short discussions among the participants. I will also be available for one-to-one counselling. A selection of texts will be sent to the participants, in case they want to prepare for the workshop. However, reading these is not mandatory.
 
Maximum number of attendees: 30
 
Full day workshop (2 x 3 hours, ideally over two days)
 

Jochem Tolsma, Rob Franken, Anne Maaike Mulders, Daniel Cowen

 
In this workshop you will collect and then describe academic departmental co-publishing networks. You will learn to webscrape scientific metadata of scientific university and departmental websites (via R packages like rvest and RSelenium), assign name-based gender and ethnicity signals, retrieve scholars’ publications (via e.g. Google Scholar and OpenAlex), and construct (longitudinal) co-publishing networks.
 
In our description of these networks, we will focus on the degree of segregation. You will answer questions such as: Are co-publishing networks segregated by scientific success and/or gender? And is segregation in co-publishing networks related to university/departmental characteristics?
 
For each step, we provide clear (proof-of-principle) coding examples and output data, ensuring you will not get stuck along the way. Depending on your skills and progress, you might want to collect and describe your own chosen universities or departments.
 
This workshop can be followed as a standalone workshop but in our second workshop ‘Analyzing the Structure and Evolution of Scientific Co-publishing Networks’ we will describe and analyze the same type of webscraped co-publishing network data in more detail by employing RSiena.
 
You will keep track of your work via a labjournal on GitHub. We will help you setting up this labjournal in the week(s) before the workshop.
 
Prerequisites:
 
Intermediate familiarity with working in R and using R Markdown
 
A beginner’s understanding of SNA
 
Entry-level of git, and GitHub
 

Tuesday, 25th June AM Half Day

Pavel N. Krivitsky, David R. Hunter

 

Description: This workshop will provide a tutorial of advanced usage of ‘ergm’ and extension packages, focusing on binary networks. Topics include specifying complex structural constraints, estimation tuning, representing complex effects with term operators, and observational (e.g., missing data) structure. Also included is using the new ‘ergm.multi’ package for modelling multi-layer and multi-mode networks, as well as joint models for ensembles of networks. Statnet is an open source collection of integrated packages for the R statistical computing environment that support the representation, manipulation, visualisation, modelling, simulation, and analysis of network data.
 
Prerequisites: Familiarity with R and ‘ergm’ required. If you are new to ERGMs, the introductory workshop on ERGMs using Statnet is strongly suggested.
 
Length: 3 hours
 
Capacity: 30
 

George Tzougas, Alex Jose

Neural networks have gained relevance in graph theory for tasks such as node classification and link prediction. Graph neural networks (GNNs) leverage the underlying graph structure to enhance the learning process, making them effective for analyzing interconnected data. Their ability to capture complex relationships in graphs has applications in social network analysis, recommendation systems, and biological network modelling.

 

In this Neural Networks Workshop, we’ll delve into the fundamentals of neural networks, powerful machine learning models inspired by the human brain. These networks excel in processing both structured and unstructured or mixed-type data. For example, in finance, neural networks can predict stock prices using structured financial data or analyze sentiments from unstructured news articles.

 

The workshop will guide you through implementing neural networks in R, specifically with the `keras` package. Practical hints will be provided for optimizing model architecture, handling data pre-processing, and interpreting results effectively. Join us to understand the core concepts of neural networks and their application in domains like finance and insurance, utilizing the capabilities of R and the `keras` package.

 

 

Filip Agneessens

 

Negative ties (such as dislike, conflict, bullying and avoidance) are increasingly being studied (Harrigan, Labianca, & Agneessens, 2020). However, classic measures applied to positive (and neutral) ties do not generally work well for negative ties (Everett & Borgatti, 2014).
 
In this 3-hour workshop we explore specific measures and approaches to study negative and signed networks. We discuss how to measure negative ties, differentiate between different types of signed networks, explore centrality measures for negative and signed networks (including the PII and PN index), explore ways to measure “cohesion” for negative and signed networks, discuss ways to study balance theory, and touch upon structural equivalence and blockmodeling.
 
This is a 3 hour workshop. Maximum number of participants: 100
 

Alberto Caimo

 

Bayesian analysis is a promising approach to social network analysis because it yields a rich fully probabilistic picture of uncertainty which is essential when dealing with relational data. Using a Bayesian framework for exponential random graph models (ERGMs) leads directly to the inclusion of prior information about the network effects and provides access to the uncertainties by evaluating the posterior distribution of the parameters. The growing interest in Bayesian ERGMs can be attributed to the development of very efficient computational tools developed over the last decade.
 
This hands-on workshop will provide participants with the opportunity to acquire essential knowledge of the main characteristics of Bayesian ERGMs using the Bergm package for R.
 
Topics will include:
– Brief overview of ERGMs;
– Introduction to the Bayesian analysis for networks;
– Prior specification;
– Model fitting and model selection;
– Interpretation of model and parameter posterior estimates;
– Model assessment via goodness-of-fit procedures.
 
The workshop will have a strong focus on the practical implementation features of the software that will be described by the analysis of real network data.
 
Interactive material will support the acquisition of concepts and understanding of the tutorial through code, scripts, and documentation.
 
PREREQUISITES: Basic knowledge of social network analysis and R. Participants are recommended to bring a laptop with R/RStudio, and Bergm installed.
 
Bergm on CRAN: https://CRAN.R-project.org/package=Bergm
 
WEBSITE: http://acaimo.github.io/Bergm/
 
REFERENCES
– Caimo, A., Bouranis, L., Krause, R., and Friel, N. (2022) “Statistical Network Analysis
 
with Bergm.” Journal of Statistical Software, 104(1), 1–23.
– Caimo, A. and Friel, N. (2014). Bergm: Bayesian Exponential Random Graphs in R, Journal of Statistical Software, 61(2), 1-25.
 
INSTRUCTOR: Alberto Caimo, University College Dublin, Ireland
 

Robert Krause, Tomáš Diviák

 

 

This 3-hour workshop provides an introduction to statistical methods for analyzing social networks. The focus is on nodal and dyadic level analysis. We will be using R packages xUCINET, migraph, and sna to perform these analyses.

 

The course outline is as follows:

 

1) testing a network’s basic properties using conditional uniform graph (CUG) test (e.g., reciprocity, homophily)

 

2) nodal level statistical tests

 

3) permutation-based comparisons between groups of nodes

 

4) QAP correlation and linear regression – the underlying logic of QAP, data format etc.

 

5) QAP GLM – logistic, poisson, and other types and extensions

 

Tom A.B. Snijders, Johan H. Koskinen

 

 

Multilevel network analysis comes in two flavors: multilevel analysis of networks (‘MAN’) where the data consists of a set of multiple networks which are conceptually similar but have disjoint node sets and no connections between them, and regarded as independent replications with respect to the social processes investigated; and analysis of multilevel networks (‘AMN’) defined as multiple interdependent networks with several node sets, some of which are shared; e.g., a one-mode friendship network between individuals together with a two-mode network of the activities of the same set of individuals.
 
This workshop is about the first kind (MAN), in particular the random coefficient multilevel longitudinal network analysis implemented in the function sienaBayes implemented in multiSiena, the sister package of RSiena. This method is based on the Stochastic Actor-oriented Model (`SAOM’). The basic idea of this random coefficient model will be presented, with the approach taken by the analysis using sienaBayes. The use of this function will be explained, and guidance will be given for parameter interpretation.
 
Topics treated are:
 
principles of Bayesian inference; the random coefficient multilevel version of the SAOM (`ML-SAOM’); MCMC estimation of the ML-SAOM; operation of sienaBayes; parameter interpretation.
 
Prerequisites
 
The workshop is intended for participants who know about the Stochastic Actor-oriented Model, and have practical experience in working with RSiena.
 
Literature:
 
Ripley, Ruth M., Tom A.B. Snijders, Zsofia Boda, András Vörös, and Paulina Preciado (2023). Manual for RSiena.
 
URL: https://www.stats.ox.ac.uk/~snijders/siena/RSiena_Manual.pdf
 
Koskinen, Johan H. and Tom A.B. Snijders (2023). Multilevel longitudinal analysis of social networks. Journal of the Royal Statistical Society, Series A.
 
DOI: https://doi.org/10.1093/jrsssa/qnac009
 
SIENA website: http://www.stats.ox.ac.uk/~snijders/siena
 
This is the first of a pair of connected 3-hour workshops. Participants are free to choose one or both of them.

Maximum number of participants 30.

Christian Steglich, Alla Loseva

 

How much can observed network distances between ethnic groups be explained by in-group clustering, or by out-group avoidance? To what degree does performance homogeneity in advice seeking networks result from selecting advisors, or from being influenced by them? How much does it matter for innovation adoption when a network is rewired while a diffusion process unfolds on it?
 
These questions seek to explain an emergent macro-level outcome while invoking competing micro-level explanatory mechanisms. In the presence of information-rich social network data, this becomes a task of statistical quantification.
 
Suitable tools for such a quantification exercise are, in principle, the stochastic network models commonly used for empirical analysis: exponential random graph models and stochastic actor-oriented models. Usually used for testing behavioural (micro-level) theories of social action in an interdependent social network setting, these models are computationally complex. Their calibration to empirical network data is achieved by means of simulation-based statistical inference. For studying emergent macro-phenomena, we can repurpose this simulation framework.
 
In this workshop, important considerations for design and implementation of such simulation studies are discussed. Hands-on illustrations, navigating between calibration and simulation, are based on the R-packages ‘RSiena’ and ‘ergm’.
 
 
We go for a 3 hour workshop with max 30 participants
 

Anni Hong, Nynke M.D. Niezink

 

Social actors are often embedded in multiple social networks, and there is a growing interest in studying social systems from a multiplex network perspective. Consequently, there is a growing demand for analytical methods and tools for these network structures. This workshop offers a practical introduction to the multiP2 R package for analyzing multiplex network data. Participants will learn the essentials of our Bayesian multiplex mixed-effects network model in the p2 (van Duijn et al., 2004) modeling framework and gain hands-on experience with the entire workflow, from data wrangling to model interpretation and assessment through a data example. The workshop will enable participants to model cross-layer dyadic dependencies as fixed effects and actor-specific dependencies as random effects, while also considering the influence of covariates in the analysis of cross-sectional, directed binary multiplex network data.
 
topics includes:
– Introduction to the multiplex p2 modeling framework
– a brief introduction to Bayesian analysis
– Overview of the R package multiP2 and the underlying estimation procedure in stan
– Data preparation
– Picking priors via prior predictive checks
– Model fitting and convergence diagnostics
– Interpretation of model coefficients
– Goodness-of-fit assessment via simulations and plotting
 
Note: participants are expected to have a basic familiarity with R for the practical segment of the workshop and some understanding of statistical inference for the conceptual portion.
 
Expected length: 3 hr, Max attendance: 20
 

Katherine Ognyanova

 

This workshop will cover network visualization using the R language for statistical computing (cran.r-project.org) and RStudio (posit.co). Participants should have some prior knowledge of R and network concepts.
 
The workshop will provide a step-by-step guide describing through series of examples the path from raw data to graph visualization in the igraph and Statnet frameworks.The advanced portion of the workshop will touch on dynamic visualization for longitudinal networks and combining networks with geographic maps. We will also discuss ways of converting graphs in R to interactive JavaScript-based visualizations for the Web.
 

Tuesday, 25th June PM Half Day

Tom A.B. Snijders, Johan H. Koskinen

 

This workshop is a continuation of the workshop “Introduction to the analysis of multilevel network dynamics using multiSiena”, treating further details.
 
It is about the random coefficient multilevel longitudinal network analysis implemented in the function sienaBayes implemented in multiSiena, the sister package of RSiena. This method is based on the Stochastic Actor-oriented Model (`SAOM’).
 
Topics treated are:
 
model specification, in particular the choice for effects whether they are randomly varying between groups or group-constant; plotting results; checking convergence; posterior means for groups; and an extended example.
 
Prerequisites
 
The workshop is intended for participants who have followed the workshop, “Introduction to the analysis of multilevel network dynamics using multiSiena”, or equivalent knowledge of the function sienaBayes.
 
Literature:
 
Ripley, Ruth M., Tom A.B. Snijders, Zsofia Boda, András Vörös, and Paulina Preciado (2023). Manual for RSiena.
 
URL: https://www.stats.ox.ac.uk/~snijders/siena/RSiena_Manual.pdf
 
Koskinen, Johan H. and Tom A.B. Snijders (2023). Multilevel longitudinal analysis of social networks. Journal of the Royal Statistical Society, Series A.
 
DOI: https://doi.org/10.1093/jrsssa/qnac009
 
SIENA website: http://www.stats.ox.ac.uk/~snijders/siena
 
This is the second of a pair of connected 3-hour workshops. Participants are free to choose one or both of them.
Maximum number of participants 30.
 

Nynke M.D. Niezink

 

Stochastic actor-oriented models (SAOMs), implemented in the R package RSiena, are widely used to study social network dynamics and the processes governing the interdependent dynamics of networks and individual behavior, such as social selection and peer influence (also known as contagion or diffusion). In the original SAOM, individual behavior variables were assumed to be measured on an ordinal categorical scale.
 
Many individual variables of interest, such as performance or health-related measures, are more naturally measured on a continuous scale. This workshop therefore introduces the SAOM for the dynamics of networks and continuous actor behavior. During the workshop, we will first introduce stochastic differential equations, the modeling framework used for dynamic continuous behavior variables. We will also discuss parameter interpretation, explained variance, and goodness-of-fit. The workshop combines a lecture with hands-on exercises using the RSiena package in R.
 
Note: You need basic familiarity with R and SAOMs to benefit from this workshop. If you have not used RSiena before, consider taking the introductory RSiena workshop.
 
Length of the workshop: 3 hours
Maximum number of attendees: 30
 

Srebrenka Letina, Mark McCann

 

The network approach is increasingly employed to explore relationships among concepts, specifically the relationships between co-occurring health conditions (e.g., using binary hospital stay data) and the relationships between psychological variables (continuous survey data). Given the multitude of approaches available for constructing and analyzing such networks and their application across different fields of study, determining the most appropriate methods and analyses can be challenging.
 
In this workshop, we aim to provide:
 
Theoretical Framework: An overview of the theoretical basis for applying network analysis to study relationships among health conditions or individual attributes.
 
Methodological Approaches: An exploration of existing methodologies for constructing networks and robustness testing of their estimations.
 
Analytical Techniques: A comprehensive set of analyses applicable to co-occurrence or correlation networks, including basic descriptive analysis, filtering methods, community detection, centrality analysis, network comparisons, motif analysis, and more.
 
We will offer a critical assessment of methods tailored to specific types of data and interpretations. Practical demonstrations of the methods and analyses will be conducted using several datasets and different R packages. In the final segment of the workshop, participants are encouraged to discuss the application of these methods to their specific datasets.
 
Length of the workshop: 3 hours
 
Maximum number of attendees: 30
 

 

David Tindall

 

This workshop is intended for relative newcomers to social network analysis. The workshop will provide an introduction to social network data collection with an emphasis on social survey methods. The workshop will consider a variety of related methodological issues such as research design, measurement, sampling, data analysis, and ethics, as well as the linkage of these issues to data collection. Different types of data collection techniques will be illustrated such as the name generator, position generator, and name roster. The different opportunities and constraints associated with data collection for whole versus ego-networks will be considered. Some discussion of non-survey techniques may also be provided. Some attention may also be given to mixed methods.
 

Michelle Birkett, Patrick Janulis, Gregory Phillips, Kate Banner, Joshua Melville, Bernie Hogan

 

This workshop will provide an orientation for conducting network data capture for public health within Network Canvas. Network Canvas (http://www.networkcanvas.com) is a free and open-source software suite that facilitates the collection of self-reported social network data. It uses touch-optimized interfaces in an interviewer-assisted environment.

 
In this workshop, we will provide an orientation to using Architect, Network Canvas’s visual interview builder, as well as Interviewer, which is used in the field to collect data directly from participants. We will also provide an orientation to data export and analysis. Our examples will draw specifically from public health.
 
This workshop will provide a comprehensive introduction to implementing research projects in the Network Canvas suite. Expect the opportunity to engage in hands-on exercises during the session with assistance from our team. When completed, you will acquire the skills to:
– Design an egocentric or whole network public health survey
– Deploy and manage a study
– Obtain data from field devices, and export it for analysis
 
Attendees should have a basic understanding of social network data capture.
 
Participants should download and install the Network Canvas Architect and Interviewer apps from the project website (https://networkcanvas.com/download.html) prior to the workshop.
– Essential: Participants must bring their own device that runs Windows 10 (1909) or newer, or macOS 10.15.4 or newer, or Linux.
– Optional: Participants are welcome to bring additional devices (tablet computers running Android 10 or newer, and iPadOS 14.2 or newer, or select models of Chromebook) to explore the software on other systems or devices.
– Attendees are encouraged to watch our 7-minute project overview video to gain familiarity of the project and tool: https://www.youtube.com/watch?v=XzfE6j-LnII
 

Cornelius Fritz, Michael Schweinberger

In social networks, ties depend on other ties, owing to social processes that give rise to transitive closure and other structural mechanisms. While in small networks, ties may depend on all other ties, this is not plausible for large networks encompassing more than a few hundred actors as social networks are more local than global. Exponential Random Graph Models (ERGMs) with local dependencies in the form of subpopulations are a model class that respects the local nature of social networks by assuming that ties only depend on other ties within the same subpopulation. If the subpopulations are observed, the network is a special case of a multilevel network. Otherwise, the subpopulations need to be learned from data.
 
The proposed workshop focuses on these types of next-generation ERGMs of small and large networks with up to 200,000 actors using the R package bigergm, which helps
– uncover who is close to whom;
– learn the local forces that govern ties among actors who are close to each other.
 
The basic ideas of next-generation ERGMs will be introduced and demonstrated by examples. Participants will be provided with sample R scripts and lecture slides.
The workshop will include a hands-on introduction to the novel R package bigergm, which is a large-scale version of the R package hergm.
 
References
 
Michael Schweinberger and Mark Handcock (2015). Local dependence in random graph models: Characterization, properties and statistical inference. Journal of the Royal Statistical Society, Series B (Statistical Methodology). 77, 647-676.
 

Thomas W Valente

 

This workshop introduces the many ways that social networks influence individual and network-level behaviors. It also provides a brief introduction to analytic approaches for understanding network influences on behaviors; and reviews existing evidence for the utility of using social network data for behavior change in a variety of settings including health behaviors and organizational performance. The workshop presents a typology of network interventions and reviews existing evidence on the effectiveness of network interventions. (Students familiar with the R environment may follow an R script written to demonstrate the 24 or so tactical interventions presented.) No software or computing requirements are needed. The workshop will be conducted by Tom Valente who has been developing and implementing network-based interventions for nearly 25 years.
 
No pre-requisites.
 
½ session (3 hours)
 
No limit on the number of attendees
 
 

Örjan Bodin, Manuel Fischer

 

In this workshop we will elaborate on how coupled social-ecological systems (or coupled natural and human systems) have been described and analyzed as multilevel networks and the research questions that have been addressed. Further, we will take stock in recent research that has identified different possibilities and barriers for further developments of this line of research.
 
Critical issues such as what are nodes and links in a social-ecological system and how to accomplish some level of comparability across different study contexts will be addressed.
 
We will also discuss the range of problems (design, data collection, methodological) that many have encountered when doing this kind of synthetic research.
 
In addition, there will be practical hands-on exercises on how conduct and understand analytical results deriving from multilevel network analyses. The analyses will be utilizing the MPNet software (http://www.melnet.org.au/pnet), which should be downloaded and installed prior to the workshop. Since MPnet require Windows, an alternative software is Statnet (https://statnet.org/), although using Statnet, not all of the multilevel analyses will be possible to conduct.
 
All exercises and examples will be based on real data, and both patterns of social relations among actors as well as environmental interactions among biophysical components will be examined. The workshop includes the following elements:
 
1. Why a social-ecological network approach? What are the presumed benefits?
 
2. What is a node, and what is a link in a complex social-ecological system?
 
3. How to move beyond just describing a social-ecological system as a multilevel network to actually ask some challenging questions, and perhaps even get some answers?
 
4. Investigate how patterns of social- and social-ecological relations among resource users can be related to social- and environmental outcomes.
 
5. Gain exposure to commonly used software for studying multilevel social-ecological networks, i.e. multilevel ERGMs implemented in MPnet.
 
Prerequisites
 
Familiarity with the concept of networks (i.e. nodes and ties) as well as some experiences of network-centric analyses. Previous exposure to ERGM is valuable.
 
3 hours workshop, maximum 30 people
 

Pavel N. Krivitsky, Carter T. Butts

 

This workshop provides instruction on how to model social networks with ties that have weights (e.g., counts of interactions) or are ranks (i.e., each actor ranks the others according to some criterion). We will cover the use of latent space models and exponential-family random graph models (ERGMs) generalised to valued ties, emphasising a hands-on approach to fitting these models to empirical data using the ‘ergm’ and ‘latentnet’ packages in Statnet. Statnet is an open source collection of integrated packages for the R statistical computing environment that support the representation, manipulation, visualisation, modelling, simulation, and analysis of network data.

 
Prerequisites: Familiarity with R and ‘ergm’ required. If you are new to ERGMs, the introductory workshop on ERGMs using Statnet is strongly suggested.
 
Length: 3 hours
 
Capacity: 30
 

Tuesday, 25th June Full Day (with lunch break)

Jochem Tolsma, Rob Franken

 

In this workshop you will analyze academic departmental co-publishing networks. You will learn to analyze the structure and evolution of these networks by means of RSiena.
 
We will start with a gentle introduction to RSiena. We will discuss the basic logic and core assumptions and provide you with a recommended workflow. In no time you will learn how to prepare your data, estimate a basic model and assess the goodness of fit.
 
Next, we will discuss many of the common structural, covariate and behavioral effects (and the subtle differences between them) and when (not) to include these in your model.
 
During the workshop you will answer questions such as: are co-publishing networks segregated by scientific success and gender? And is segregation in co-publishing networks the result of structural network effects, influence processes and selection processes?
 
For each step, we provide clear (proof-of-principle) coding examples and output data, ensuring you will not get stuck along the way. Depending on skills and progress, we are more than happy to introduce you to, and assist you with, the more intermediate and advanced feautures of RSiena.
 
This workshop can be followed as a standalone workshop but in our first workshop ‘Webscraping Scientific Co-publishing Networks’ you will learn the necessary skills to webscrape co-publishing network data yourself.
 
You will keep track of your work via a labjournal on GitHub.We will help you setting up this labjournal in the week(s) before the workshop.
 
Prerequisites:
 
Intermediate familiarity with working in R (base and tidyverse) and R Markdown,
 
A beginner’s understanding of SNA via stochastic actor-orientated models
 
Entry-level of git, and GitHub
 

Elisa Bellotti, Betina Hollstein

 

The workshop focuses on the use of mixed methods research designs when studying whole and ego-centered social networks. The workshop will be conducted in two parts. The first part introduces social network qualitative research and the principles of mixed methods research designs and its contributions to the study of social networks, pointing out advantages and challenges of this approach. Illustrations of the theoretical and methodological aspects are given by bringing examples from a variety of fields of research. The second part is devoted to the presentation of concrete procedures to apply mixed methods in network research both at the level of data collection and analysis. This part includes an introduction of different approaches to the collection of whole and ego-centered network data, i.e. interviews, ethnographic methods, archival data, together with visual instruments. It then moves to the analysis of the quantitative and qualitative dimensions of network relationships and structures in a mixed method perspective.

 

Juergen Lerner, Alessandro Lomi

 

Networks of social relations and communication networks frequently generate information on repeated interaction over time. This information typically takes the form of relational event sequences – streams of time-ordered events connecting social actors. Examples of relational events are common. Conversations, email communication, interaction among members of teams, participation in social gatherings or in peer-production projects, are all examples of interactive settings that may generate observable streams of relational events. In this workshop we will specifically discuss “polyadic” social interaction processes in which events can connect varying and potentially large numbers of actors simultaneously. Examples of such polyadic events (or “hyperevents”) include sequences of meeting events or social gatherings, connecting all of their participants simultaneously, or multicast (i.e., “one-to-many”) communication events such as emails in which one actor sends a message to several receivers.
 
The first part of this workshop (“beginners”) provides a hands-on introduction to relational hyperevent models (RHEM). We start with a quick overview of empirical settings and research questions that can be addressed by a RHEM analysis, introduce the basic model, and illustrate practical analysis of the famous Davis, Gardner, and Gardner “Deep South / Southern Women” data with the open-source software eventnet (https://github.com/juergenlerner/eventnet).
 
The second part of this workshop (“advanced”) discusses several variations of the basic model. RHEM can be applied in various application settings, each comming with possibly different structural constraints, network effects, challenges, and opportunities. The application scenarios that we discuss include:
– directed multicast communication networks (e.g., Enron email corpus)
– events representing team-work that has an outcome (e.g., coauthorship/publication networks; competition in team-sports)
– citation networks (e.g., scientific papers; patents); co-evolution of team-formation and citations to prior work
– networks of labeled hyperevents (e.g., meetings associated with one or several “purposes” or “topics”)
 
Participants will learn a broad set of techniques and variations that enable them to apply RHEM in their own research projects.
 
The workshop is targeted at participants interested in statistical modeling of networks based on relational event data – with a specific focus on polyadic, multicast, or one-to-many interaction events. Participation in the first part (“beginners”) does not assume any particular prior knowledge or experience with statistical models for social networks. Participation in the second part (“advanced”) assumes participation in the first part or, alternatively, basic knowledge about statistical models for networks, such as REM, ERGM, SAOM/Siena, or similar models. It is possible to attend both parts or only one of the two parts, depending on interest and prior knowledge.
 
Participants are invited to point us to their own research projects, which may possibly be addressed by a RHEM analysis, prior to the workshop.
 
For additional information on the type of models discussed, we suggest the following references:
– Lerner and Lomi (2022). A dynamic model for the mutual constitution of individuals and events. Journal of Complex Networks, 10(2):cnac004. https://doi.org/10.1093/comnet/cnac004
– Bianchi, Filippi-Mazzola, Lomi, and Wit (2023). Relational Event Modeling. Annual Review of Statistics and Its Application.
 
as well as the software tutorials and further papers linked from here: https://github.com/juergenlerner/eventnet/wiki
 

 

Andrea Fronzetti Colladon, Roberto Vestrelli, Francesca Grippa

 

Leveraging the power of big data represents an opportunity for researchers and managers to reveal patterns and trends in social behaviors and consumer perceptions. This workshop shows how to successfully integrate Text Mining with Social Network Analysis for business and research. It presents the Semantic Brand Score (SBS) and other powerful methods and tools for analyzing semantic networks, studying brand/semantic importance, and performing advanced NLP tasks. The workshop also describes the functionalities of the SBS Business Intelligence App (SBS BI), designed to produce a wide range of analytics and mine textual data. We discuss several case studies and show how these methods have been used, for example, to predict tourism trends, select advertising campaign testimonials, or make economic, financial, and political forecasts. SBS BI analytical power extends beyond “brands”, comprising applications to study: commercial brands (e.g., Pepsi vs. Coke); products (e.g., pasta vs. pizza); personal brands (e.g., name and image of political candidates); set of words representing values (e.g., a company’s core values) or concepts related to societal trends (e.g., terms used in media communication that impact consumers’ feeling about the state of the economy). Combining text analysis with network science can change how we make decisions and manage organizations in the era of big data.
 
More info and materials are available at: https://learn.semanticbrandscore.com
 
Length of the workshop: 6 hours.
 
Maximum number of attendees: 50 (but can be flexible)

Viviana Amati, Marion Hoffman

 

This workshop offers a basic introduction to the theory and application of Stochastic Actor-oriented Models (SAOMs). SAOMs are a statistical model family developed for the analysis of social networks panel data, understood here as two or more repeated observations of a network on a given node set (usually between twenty and a few hundred nodes). The method is implemented in the RSiena, package in the R software.
 
The first part of the workshop will focus on the intuitive understanding of the model and operation of the software. The second part will present models for the simultaneous dynamics of networks and behavior and other more advanced topics such as model specification, multivariate networks, and goodness of fit checking.
 
Course participants should have a basic understanding of social network analysis concepts and methods and basic knowledge of the R programming language is necessary to successfully follow the workshop. Basic knowledge of multivariate statistical models (e.g. linear regression) is recommended. They should bring a laptop to the workshop with the latest versions of R, RStudio (or their preferred GUI if any) and the RSiena R package installed.
 
Names and contact information of all organisers:
 
Viviana Amati
 
Department of Statistics and Quantitative Methods, University of Milano-Bicocca, Milan, Italy
 
Marion Hoffman
 
Institute for Advanced Study in Toulouse (IAST), Toulouse, France
 
Department of Sociology, University of Zurich, Zurich, Switzerland
 
Length of the workshop: 6 hours
 
Maximum number of attendees: 20
 
CONTACT
Heriot-Watt University, Riccarton, Edinburgh, EH14 4AS