All time references are in CEST
Survey and Sampling Methodologies in Network Data |
|
| Session Organisers | Dr Francesco Santelli (University of Trieste) Dr Gianpaolo Caprino (University of Milano-Bicocca) Professor Domenico De Stefano (University of Trieste) Professor Susanna Zaccarin (University of Trieste) Professor Rosario D'Agata (University of Catania) Dr Viviana Amati (University of Milano-Bicocca) |
| Time | Tuesday 18 July, 09:00 - 10:30 |
| Room | |
The interest in relational data has introduced new opportunities and challenges for survey research. This session explores the methodological complexities of collecting and sampling network data encoding interactions among entities. As researchers increasingly rely on unconventional and multiple data sources, such as bibliometric databases, social media platforms’ interactions, IoT, and sensor data, adapting traditional survey methodologies becomes paramount, especially without sufficient attribute information.
This session aims to spark a discussion on the future of data collection in network research. We welcome contributions on
- The development of adequate instruments measuring interactions in cross-sectional and longitudinal studies
- The advancement of approaches dealing with inherent biases that arise from, e.g., non-random snowball, dyadic, and ego-network sampling structures
- The progress in strategies to obtain representative network samples
- The use and integration of primary and secondary data
- Ethics in network data collection
- The development of software and tools for data collection.
Keywords: Collection of relational data, Ethics, Network Sampling, Network surveys, Software
Professor Marian-Gabriel Hâncean (University of Bucharest) - Presenting Author
Professor Jürgen Lerner (University of Konstanz)
Personal Network Analysis (PNA) affords a robust framework for examining
individuals' social embeddedness by capturing relational (ego-alter and alter-alter
ties) and attribute data (e.g., socio-demographics, perceptions). Despite its analytical
power, most PNA studies are cross-sectional, limiting their capacity to observe
changes over time. Longitudinal panel studies remain rare, with few notable
examples.
We address methodological challenges encountered in collecting panel personal
network data in a rural Romanian community (N = 4,124). Data were collected
across two waves (n = 83 egos, 1,970 alters in September 2023; n = 94 egos, 1,513
alters in March 2024), with 68 egos participating in both waves. To reduce
respondent cognitive burden, we transitioned from a fixed-name generator (n = 25
alters) in the first wave to a free-name generator in the second wave (avg = 16
alters). Alters were tracked across waves using unique identification codes, enabling
analysis of changes in alter characteristics and relationships. For inconsistently
identified alters, we applied data-cleaning techniques by cross-referencing
demographic and relational attributes (e.g., age, gender, relationship type) without
employing imputation methods.
The study employed link-tracing sampling, initiated from six seeds, with a non-fixed
number of recommendations per study participant. Data collection is planned over
four waves, concluding in September 2025, to evaluate community-level public
health interventions aimed at cancer prevention—an underexplored area in rural
Eastern Europe.
We aim to: (1) identify and discuss methodological challenges in collecting panel
PNA data using link-tracing sampling and name generators, and (2) evaluate their
impact on the structural and compositional attributes of personal networks. We
address challenges in modeling network data, including disentangling social
selection, influence, and contextual effects.
Our insights contribute to refining network-based sampling techniques and
advancing panel PNA methodologies, particularly in underrepresented rural contexts.
Dr Philip Adebahr-Maskow (Martin Luther University Halle-Wittenberg (MLU)(MLU)) - Presenting Author
Miss Theresia Ell (GESIS – Leibniz Institute for the Social Sciences (Mannheim))
Dr Lydia Repke (GESIS – Leibniz Institute for the Social Sciences (Mannheim))
In cross-sectional and longitudinal studies, negative ties are often measured using the identification of difficult people as a single-item indicator. This widely used name generator elicits negative members of egocentric social networks. Alternative methods also exist, such as asking respondents about individuals with whom they experience conflict, feel worried, or find annoying. Each captures different facets of negative relationships. It is crucial to understand what the single-item measure of difficult people captures, its overlap with other measures, and potential biases. For instance, this generator may overemphasize certain types of negative relationships or distort perceptions of negative ties.
To address this, we evaluate the validity of the difficult people network generator by comparing it with three additional items focused on conflict, worry, and annoyance.
Our study uses data from the LoneCovid project, which investigates changes in social contacts during the COVID-19 pandemic. Data were collected from a subsample of the GESIS Panel, a mixed-mode panel of approximately 5,200 individuals from the German population. A total of 118 participants generated egocentric networks of 20 alters each during live video-based interviews in 2023 and 2024.Our analysis includes descriptive comparisons to identify biases in the single-item measure of difficult people versus the three alternative measures, examining variations by participant and relationship characteristics (e.g., age, gender, education, relationship type, tie strength). We also conduct exploratory factor analysis (EFA) to assess convergent validity and compare predictive validity by examining how well each measure predicts external variables.
The findings reveal biases in the difficult people generator and offer insights into its internal validity. These results enhance the understanding and quality of negative tie data collection in social network studies, contributing to more robust future research designs.
Mr Roberto Casaluce (University of Catania) - Presenting Author
Professor Rosario Giuseppe D'Agata (University of Catania)
We present MaScoNet, a web-based application designed to collect ego-centered network data through a structured questionnaire, tailored to gather scientific collaboration data. MaScoNet enables researchers to examine how informal and formal forms of scientific collaboration influence scholars’ performance.
When collecting such data, it is necessary to identify both the focal individuals (egos) and their associated network members (alters). Alters can be identified by using name-generator questions—where participants list their collaborators—or by having the researcher pre-select alters. Both methods may introduce biases, such as recall bias or researcher subjectivity, and MaScoNet offers flexibility to choose the most suitable approach based on study objectives.
The questionnaire in MaScoNet is structured into two parts. The first part consists of questions designed to gather data on the types of co-authorship relationships, both formal and informal, that have led to producing scientific outputs between the respondent and each of their co-authors. For each selected co-author, the respondent is presented with a separate page containing these questions. The second part assesses the extent of informal relationships with colleagues who have not co-authored any publications.
Furthermore, in the first section of MaScoNet, there is a feature that allows the researcher to provide a list of scientific articles co-authored by the respondent and each identified co-author. By incorporating tailored questions based on these articles, the researcher can analyze a two-mode network that complements the ego-alter network. In this two-mode network, one set of nodes represents the scientific articles, while the other set represents their authors. If multiple articles are listed, the respondent can select more than one article when answering a particular question. This feature offers deeper insights into collaborative structures and their influence on scholarly performance.