This project hosts materials presented at the [2023 ICPS convention][1], in a methodological symposium titled "New Tools to Analyze Round-Robin Data from Social Networks", chaired by Terrence D. Jorgensen. Below the general **Abstract** and **Supporting Summary** for the full symposium, you can find **Individual Abstracts** for each contribution. Links to the software tools are provided below each individual abstract. Software examples are also provided with example data in the **Files** tab. ### Abstract Round-robin designs facilitate studying social phenomena from an interpersonal perspective, but their complex network structure requires sophisticated analytical models. This symposium highlights recent developments that enable analyzing (potentially multiple) network variables with freely available, user-friendly software. Each presentation demonstrates a software tool on an example data set. ### Supporting Summary Researchers increasingly study social phenomena from an interpersonal perspective, using data gathered from a round-robin design, in which each member of a group provides data about every other member. For example, each student in a classroom can indicate how much they like each other student, which yields a complex social-network structure. The social relations model (SRM) decomposes interpersonal perceptions/behaviors into random effects associated with perceivers/actors and their targets/partners, as well as relationship-specific nuance. Sampling several small round-robin groups (e.g., families, classrooms) also enables modeling of group-level variance. The primary outcomes of an SRM analysis involve the estimated covariance matrices for person-level variability, dyad-level variability, and potentially group-level differences. This symposium is a showcase of recently developed software tools that simplify the complexities of such analyses using freely available, user-friendly software. All presentations and software examples are provided on the Open Science Framework (OSF). The first presentation discusses how to study interdependence within a family. When dyadic data are obtained for a specific family using a round-robin design, we can calculate the person- and relationship-level effects using an ANOVA-approach. To gain insight into the functioning of a particular family, one can compare the family-specific SRM to a norm sample and deduce whether that family has deviating scores on a particular SRM effect. Typically, one needs access to the norm sample’s SRM ANOVA scores. We introduce a user-friendly application that uses an alternative method for this assessment, that requires as input solely the population parameter estimates that are often reported in SRM family literature. The second presentation introduces an adaptation of the SRM specifically developed for “purely” dyadic data, which do not vary within a dyad. Purely dyadic data target something in common between two people (e.g., the number of meals two family members shared during the last week) or evaluate the dyad as a whole (e.g., the distance two people stand apart from each other). The Purely Dyadic Social Relations Model (PD SRM) enables researchers to decompose a purely dyadic measure into individual-, dyadic-, and family-level components. To easily perform such analyses, we developed a Shiny app called PDSRM, which is freely available online. In this talk, we illustrate both the model and the PDSRM app with data from Flemish families. The third presentation discusses an R package that facilitates a 2-step estimation approach for round-robin data with indistinguishable dyads (i.e., without particular family roles). The first step estimates a multivariate SRM, including covariances with dyad- and person-level covariates, and optionally with group-level covariates when multiple round-robin groups are observed. The second step uses the R package lavaan to fit an SEM to the level-specific covariance matrices. Similar to accounting for ordinal data by first estimating (thresholds and) polychoric correlations among hypothesized latent-response variables, the 2-step SR-SEM approach uses weighted least squares estimation to account for uncertainty about the estimated level-specific covariance matrices. Implementing Step 2 provides all the benefits standard SEM software provide. A new R package implementing the 2-step approach is demonstrated on social-mimicry data. ## Individual Abstracts & Materials #### *The Purely Dyadic Social Relations Model (PD SRM)* - **Authors**: Lara Stas, Leila Van Imschoot, William Cook, Tom Loeys, & Ann Buysse - **Abstract**: Sharing family meals is considered to be one of the most common and everyday family activities, but who is responsible for having these meals together? In this presentation, we show how to easily disentangle such purely dyadic variables into individual, dyadic, and family components using the PD SRM app. - [Link to Shiny app][2] #### *A practical tool for family assessment based on the social relations model* - **Authors**: Justine Loncke, Tom Loeys, & Marieke Fonteyn - **Abstract**: Using relationship-anxiety data described by [Kenny, Kashy & Cook (2006)][3], we demonstrate how a family therapist can use the SRM-Family-Assessment app to gain insight into complex family functioning. Results can indicate possible avenues that may be worth exploring further, in order to understand the family members’ experiences. - [Link to Shiny app][4] - **Open-access tutorial article**: Loeys, T., Fonteyn, M., & Loncke, J. (2021). A practical tool for family assessment based on the Social Relations Model. *Frontiers in Psychology, 12*(699831). doi:[10.3389/fpsyg.2021.699831][5] #### *Two-stage estimation of the social-relations structural equation model (SR-SEM)* - **Authors**: Aditi Bhangale & Terrence D. Jorgensen - **Abstract**: [Salazar Kampf et al. (2018)][6] studied social mimicry in small groups of strangers who spent a few minutes becoming acquainted. They hypothesized that social mimicry during conversation mediates the link between initial impressions and subsequent liking. We reanalyze their item-level data using common factors to account for measurement error. - Two-stage ML estimation available via R package `lavaan.srm` (on [GitHub][7]) - Single-stage ML estimation available via R package `srm` (on [GitHub][8] and [CRAN][9]) - Click on the **Files** tab to access example data and R syntax. [1]: https://www.psychologicalscience.org/conventions [2]: https://larastas.shinyapps.io/PDSRM/ [3]: https://davidakenny.net/kkc/kkc.htm [4]: https://srmfamilyassessment.shinyapps.io/Zscores/ [5]: https://doi.org/10.3389/fpsyg.2021.699831 [6]: https://doi.org/10.1177/0956797617727121 [7]: https://github.com/TDJorgensen/lavaan.srm [8]: https://github.com/alexanderrobitzsch/srm [9]: https://cran.r-project.org/package=srm