Adherence to the planetary health diet index (PHDI) is associated with health benefits, but its relationship with depression, particularly the roles of dietary components and symptoms, remains poorly understood. We aimed to (1) compare PHDI component interrelationships between individuals with and without depression; (2) identify PHDI components most predictive of depression using machine learning and (3) identify bridge nodes connecting PHDI components with depressive symptoms. This cross-sectional study analysed data from 25 542 adults in the National Health and Nutrition Examination Survey 2005–2018. The PHDI was calculated from two 24-h dietary recalls. Depression was assessed with the Patient Health Questionnaire-9 with scores ≥10 indicating depression. We employed network analysis to compare dietary structures, an extreme gradient boosting (XGBoost) model interpreted using shapley additive explanations (SHAP) values to identify predictors and bridge network analysis to identify bridge nodes connecting the PHDI and depressive symptom communities. The PHDI network in the non-depressed group was more densely connected than in the depressed group. SHAP analysis of the XGBoost model identified non-starchy vegetables (PHDI2) and added sugars (PHDI14) as the most important predictors, with higher component scores associated with a lower predicted probability of depression. In the bridge network analysis, appetite problems (PHQ5) and depressed mood (PHQ2) emerged as the strongest bridge nodes between the PHDI and depressive symptom communities. A structured, plant-forward dietary pattern was associated with lower depression levels, with appetite problems and depressed mood identified as the bridge nodes between the PHDI and depressive symptom communities.

The literature on U.S. politics has firmly established an increase in polarization in Congress since the 1970s, arguing that cultural issues are especially polarizing, while economic issues remain less polarized. However, increasing disagreement over the importance of economic issues from welfare state spending to taxation raises the question whether economic discourse has slowly become more polarized, too. This paper investigates issue-level polarization based on issue attention in Congress using speech data and proposes a measurement approach that combines computational text analysis with network analysis. The approach uses speech similarity between individual representatives to construct political networks over time. It measures issue-level polarization within these networks by calculating modularity scores based on the Leiden community detection algorithm. The measurement strategy shows strong patterns of issue-level polarization. This paper makes three contributions. First, it demonstrates that polarization can be studied with text analysis combined with network data. Second, it shows that representatives in Congress increasingly talk about entirely different economic issues. Third, it provides an analytical framework to compare discursive polarization in speech with issue-level polarization.

Over recent decades, international trade has expanded significantly, increasing economic integration but also amplifying systemic vulnerabilities. Recent global shocks, such as the COVID-19 pandemic and the Russia–Ukraine conflict, have exposed the fragility of international trade systems, underscoring the need for analytical tools to enable timely, evidence-based trade policy. This paper introduces TERRA, an analytical framework created by the Italian National Institute of Statistics to support the monitoring and analysis of European trade patterns using official statistics. Using high-frequency Eurostat COMEXT data, TERRA integrates network-based representations and time series indicators to identify structural vulnerabilities, emerging concentrations, and changes in countries’ roles within trade networks. Beyond interactive monitoring, the framework supports advanced analytical extensions, including programmatic longitudinal analyses and model-based counterfactual analyses, designed to explore trade reallocation mechanisms under stylized supply shocks. The analytical potential of TERRA is shown through two case studies: electricity trade during the COVID-19 pandemic and weapons and ammunition trade during the Russia–Ukraine conflict. The findings demonstrate how network-based monitoring tools, grounded in official statistics, can effectively interpret complex trade dynamics and support evidence-based trade governance.

Suicide attempts arise from an environment of many interconnected psychosocial and other factors. Yet, the relationships between these factors have not been studied widely in low- and middle-income countries, where an estimated 73% of suicides occur. We aimed to identify the patterns and complexrelationships between the factors associated with suicide attempts. We conducted a network analysis on data from 25,761 school-going adolescents across12 low- and lower middle-income countries (LLMICs), obtained from the Global School-based Health Survey. Using Ising network analysis, we examined theinterrelationships between past-year suicide attempts and several factors; we estimated centrality measures and examined network stability usingbootstrapping. Suicide attempt prevalence was 11.8%. The network analysis identified three clusters: mental health, physical harm and behavioral factors. Suicide attempts showed strong conditional associations with drug use, bullying and worrying, while having close friends was a prominent protective factor.Behavioral factors: alcohol use, drug use and smoking were the most central nodes, showing the highest expected influence. This study provides novel insightsinto the interplay between factors associated with adolescent suicide attempts in LLMICs. Comprehensive interventions could focus on clustered risk factors,while empowering protective factors and addressing behavioral factors could conditionally attenuate other associations in the network.

How do spatial proximity, policy networks, and temporal dynamics interact to shape education policy diffusion across the American states and have these channels of influence changed over time? This study introduces an integrated framework combining spatial-X (SLX) estimation, exogenously constructed policy-pathway network exposure, and discrete-time survival models with policy-level random intercepts to decompose the relative contributions of geographic, relational, and temporal channels. Drawing on 10 education policies from the State Policy Innovation and Diffusion (SPID) database (1896–2017), with the policy-pathway network constructed from 714 non-education SPID policies to address within-domain endogeneity concerns, the analysis finds that geographic proximity and network exposure each predict adoption, although their relative importance has shifted across historical periods. For policies whose diffusion windows opened before 1970, geographic proximity is the dominant detectable channel; for policies opening after 1970, network exposure displaces geography. Spatial effects are detectable for K-12 policies but not for higher-education reforms, where statistical power is limited. A complementary spatial Durbin model on rapid-diffusion versus sequential-diffusion subsamples shows that contemporaneous (SAR) determination dominates for policies that diffuse in fewer than seven years, while temporally lagged (SLX) determination dominates for sequential-diffusion policies. A SPID-wide replication on 635 policies and 17,628 adoption events confirms both spatial signals across the broader policy landscape. These findings reveal how the institutional infrastructure connecting American states has evolved from geography-mediated to network-mediated channels, with implications for diffusion theory, methodological practice, and intergovernmental policy coordination.

The study of polarization in social networks is complicated by conceptual confusion: key forms of polarization include issue, ideological, affective, interactional, interpretive, and perceived polarization, but consensus remains elusive on which of these concepts are most important, how they are interrelated, and whether they should be considered problematic. In this article, we show how the communicative practices of the various actors in a given public debate can be analyzed to reveal the networked structures of discursive alignments and divergences and assess the degree of alliance or opposition between them, and how this analysis in turn provides a basis for the identification of polarization and its potential diagnosis as destructive. We address the challenge of assessing polarization by introducing the new analytical approach of practice mapping. Conceptually supported by practice theory, practice mapping advances beyond conventional social network analysis by no longer grounding its network analysis only in the direct network ties made by participants in social networks, but conceptualizing each participant’s broader patterns of action and interaction within the social network (including their engagement with other participants, but also their information sourcing, language choices, and other qualities of their participation) as that participant’s unique communicative practice—and then constructing and mapping the network of similarities between these practices for all participants. This serves as a key step towards identifying and analyzing patterns of polarization in public discourse. To illustrate this approach, we apply it to public posts on Facebook about the Voice to Parliament referendum in Australia. Our practice mapping was able not only to clearly identify the deeply opposed Yes and No campaigns, but also to show substantial distinctions between the highly centralized and unified No camp and the loose and uncoordinated discursive alliances supporting the Yes camp. Using this case study, we demonstrate the utility of practice mapping in assessing polarization in social networks.

This paper advances a methodological framework for the study of policy polarization. We conceptualize polarization as the structure of relationships among political and epistemic actors, measurable through the similarity of their policy statements. To generate these relational data, large language models (LLMs) are employed as scalable tools for extracting, summarizing, and synthesizing policy statements or positions from diverse textual sources, including debates, party platforms, interviews, and think tank publications. The resulting corpus is embedded into a shared vector space, and similarity measures between statements provide the basis for constructing multimodal networks that connect policy statements through semantic similarity and link them to candidates and think tanks. These networks enable the application of social network analysis, such as correspondence analysis, influence modeling, and structural comparison, to model alignments, divergences, and patterns of polarization. The framework contributes a replicable and extensible approach to analyzing policy polarization, integrating computational text processing with network-analytic models to capture the relational dynamics of political discourse.

Network analysis assessed the relationship between listening fatigue (Vanderbilt Listening Fatigue Scale; VFS) and foreign language learning anxiety (Foreign Language Classroom Anxiety Scale; FLCAS). Nodes represented the questions on each scale, and edges between nodes represented the strength and direction of the correlation between those questions/nodes. A simple correlation of the sum scores revealed a negative correlation: students reporting high fatigue reported low anxiety. This result may be due to the way in which cognitive resources are allocated (i.e., “ego depletion”). Analysis of the micro- and meso-level structure of the network revealed two distinct subsets of nodes (i.e., communities), with one community containing many of the FLCAS items (and the 5 nodes highest in strength) and the other community containing the VFS items and several items from the FLCAS that assessed confidence in using L2 rather than anxiety in learning L2. Additional ways that network science can be used to increase our understanding of language learning are discussed.

Mental disorders are major contributors to the global burden of disease and may show correlated incidence patterns across countries. We characterized the country-level ecological network of age-standardized incidence rates for 10 mental disorders, identified statistically central disorders and examined how individual countries influenced estimated network connectivity. Data for 204 countries and territories, including Sociodemographic Index values, were obtained from the Global Burden of Disease Study 2023. A regularized network model using the mgm framework estimated conditional associations among country-level variables. Strength centrality and case-deletion delta-centrality analyses assessed disorder connectedness and country influence, with bootstrap procedures evaluating stability. Approximately 400 million incident cases of the studied disorders were estimated globally in 2023. Anorexia nervosa and bipolar disorder had the highest strength centrality, indicating broader conditional connections with other disorders. Australia, the United States, Iran, China and Finland showed relatively large delta-centrality values, suggesting greater influence on estimated centrality than most countries. Strength centrality stability was acceptable but modest. These findings provide a descriptive, hypothesis-generating view of how modeled incidence rates covary across countries and should not be interpreted as individual-level comorbidity, causal influence or direct policy priority.