This project hosts materials presented at the [2023 EAM conference][1], in a symposium titled "Using Structural Equation Models to Analyze Round-Robin Data from Social Networks", which took place on Wednesday 12 July 2023 at Ghent University. The symposium was organized by Terrence D. Jorgensen and included 4 presenters:
1. **Terrence D. Jorgensen** (University of Amsterdam)
2. **Aditi Manoj Bhangale** (University of Amsterdam)
3. **Lara Stas** (Ghent University & Vrije Universiteit Brussel)
4. **Leila Van Imschoot** (Ghent University)
The abstract for the symposium is below, and abstracts for individual presentations can be found in this project's Files/Abstracts folder. Presentation slides can be found in the Files/Slides folder.
## Abstract
Study social phenomena from an interpersonal perspective is enabled by round-robin designs, in which each member of a group provides data about every other member—e.g., each student in a classroom indicates how much they like each other student, or each nuclear-family member indicates how secure their relationship is with each other family member. This complex pattern of interdependence among dyadic observations ($Y_{ij}$: a variable $Y$ measured about person $i$ responding to or interacting with person $j$) has a social-network structure, which requires sophisticated analytical models to account for interdependence. The linear social relations model (SRM: $Y_{ij}=\mu+P_i+T_j+R_{ij}$) was designed to decompose such interpersonal perceptions into random effects associated with perceivers ($P_i$), their targets ($T_j$), and relationship-specific nuances captured by dyad-level residuals ($R_{ij}$). Sampling several round-robin groups (e.g., families, classrooms) also enables modeling of group-level variance via a random intercept $\mu_g$ rather than constant mean $\mu$. A multivariate SRM can estimate correlations among multiple round-robin variables, but fitting theoretical models to explain those relationships requires a larger modeling framework, such as structural equation modeling (SEM). This symposium highlights how to analyze SRM data with SEM via open-source software. The first pair of presentations focus on family-SRM data, where family members can have differential influence within their network, modeled via factor loadings. The second pair of presentations compare and evaluate 1- and 2-stage maximum-likelihood estimation methods for analyzing multivariate-SRM data via the social-relations SEM (SR-SEM). All presentations and software examples are provided on the Open Science Framework (OSF): https://osf.io/ahuq6
This work was financially supported by NWO grants [016.Veni.195.457][2] and [406.XS.01.078][3], awarded to Terrence D. Jorgensen.
[1]: https://eam2023.ugent.be/
[2]: https://www.nwo.nl/en/projects/016veni195457
[3]: https://www.nwo.nl/en/researchprogrammes/open-competition-ssh/granted-projects
