Social Networks最新文献

The stochastic actor oriented model (SAOM) is a method for modelling social interactions and social behaviour over time. It can be used to model drivers of dynamic interactions using both exogenous covariates and endogenous network configurations, but also the co-evolution of behaviour and social interactions. In its standard implementations, it assumes that all individual have the same interaction evaluation function. This lack of heterogeneity is one of its limitations. The aim of this paper is to extend the inference framework for the SAOM to include random effects, so that the heterogeneity of individuals can be modelled more accurately.

We decompose the linear evaluation function that models the probability of forming or removing a tie from the network, in a homogeneous fixed part and a random, individual-specific part. We extend the algorithm so that the variance of the random parameters can be estimated with method of moments. Our method is applicable for the general random effect formulations. We illustrate the method with a random out-degree model and show the parameter estimation of the random components, significance tests and model evaluation. We apply the method to the Kapferer’s Tailor shop study. It is shown that a random out-degree constitutes a serious alternative to including transitivity and higher-order dependency effects.

In classical Social Network Analysis (SNA), what counted as a “social tie” was fixed by available data collection methods. The emergence of large-scale unobtrusive data collection techniques has sparked renewed interest in the very idea of what counts as a “social tie.” Importantly, there has been an acknowledgment that the core issues raised by these developments are primarily conceptual. As a result, there is renewed interest in developing a scientifically grounded characterization of what is arguably the most central concept in social network analysis. This paper contributes to this conceptual effort. I rely on a technique of conceptual representation borrowed from cognitive psychology and cognitive linguistics in which frames for concepts are represented as directed graphs linking attributes to values. I show how the frame representation helps clarify the sort of claims that network theories make (e.g., imposing restrictions on attributes and values), how it helps specify both intra and inter-conceptual relations, how it illuminates seldom noted inter-theoretical commonalities and contrasts, and how it helps avoid common conceptual pitfalls.

Adolescent health-related behaviours and outcomes are shaped by their peers through various social processes. Research using network data on friendship ties has uncovered evidence for processes such as peer influence and imitation. Much less is known about how the structure of small groups within a network, network communities that represents its meso level, affect individuals. The structure and composition of peer groups could play an important role in shaping health behaviour but knowledge of the effects of groups is limited. We used data from The Peers and Levels of Stress study, a cross-sectional social network study conducted in 2006 of 22 secondary schools in Glasgow, Scotland. Students from one year group (15–16 yrs., N = 3148; 50.8% women) provided information on socio-demographics, health behaviour and friendships via a questionnaire. Dependent variables were substance use and general mental wellbeing measured via principal components. We used a series of multilevel models with students (level 1), network communities (peer groups) identified by the Walktrap algorithm (level 2), and schools (level 3). We found substantial and moderate clustering at the peer group level for substance use and mental wellbeing, respectively. Larger and more transitive groups were associated with less substance use, but worse mental wellbeing. Addressing the methodological gap regarding the influence of the choice of group detection method on findings, we repeated our analysis using nine additional methods. The choice of the method somewhat influenced peer group variance and greatly influenced association of peer group properties with health. This study makes two key contributions to school-health improvement research. Beyond describing peer group clustering health outcomes, this is the first demonstration that structural and compositional characteristics of peer groups are associated with individual health, while highlighting the sensitivity of findings to group detection method used.

Research suggests that coming from a lower economic background compromises social integration at school, yet the precise mechanisms underlying this link remain unknown. Therefore, this study examined the effect of household income on friendship network dynamics among classmates in a large sample of Swedish youths (n = 4787 from 235 classes, m age = 14.65, 51% girls, and 33% immigrant background), using multilevel longitudinal social network analysis. Over time, students from poorer households were less often selected as a friend by classmates and they less often initiated or maintained friendship ties than students from higher income households. Furthermore, different conceptualizations of income relative to classmates did not impact friendship formation tendencies. The findings indicate that theories of relative income do not extend understanding of students’ friendship formation beyond processes related to absolute income. In addition, this study suggests that the social integration of students from low-income households could be boosted by both promoting their agency in forming friendships and preventing exclusion by classmates.

Different lines of research have argued that specific groups, such as boys or ethnic minorities, are more prone to develop an anti-school culture than others, leading to group differences in the social acceptance of high performers. Taking an ecological view, we ask to what extent the school context promotes or prevents the emergence of group-specific oppositional cultures. Theoretically, we argue that group-based oppositional cultures become more likely in schools with low socio-economic resources and in schools where socio-economic differences align with demographic attributes. We test our hypotheses based on data from a large-scale, four-wave network panel survey among more than 4000 students in Germany. Applying stochastic actor-oriented models for the coevolution of networks and behavior, we find that group-based oppositional cultures in which students like high performers less are very rare. However, in line with theoretical expectations, the less resourceful a school is, the more boys tend to evaluate high-performing peers less positively than girls do. Moreover, the more ethnic minority boys are socioeconomically disdvantaged in a school, the more they tend to evaluate high performers less positively than majority boys do.

This paper aims to understand how a group of academics cite each others’ work through time, considering the simultaneous co-evolution of three networks representing their scientific collaboration, the journals in which they publish and institutional membership. It argues that both social and cognitive processes contribute to these dynamics. Two types of network mechanisms are considered specifically: closures by affiliation and closures by association. To assess whether these mechanisms generate the macro features of the network under study, we propose new features for three-mode multilevel networks such as the mixed geodesic distances, mixed degree distributions, and the mixed quadrilateral census. We investigate whether a micro-level model that considers the above-mentioned network mechanisms is able to correctly reproduce these features. We apply stochastic actor-oriented models (SAOMs) for one-mode and two-mode networks to link the micro-macro processes using a dataset of a scientific community of astronomers from 2013 to 2015. The results suggest that social relationships grounded on scientific collaboration and proximity based on institutional affiliation are more accurately suited to understanding the co-evolution of the network of citations than an alternative approach that merely considers cognitive-based networks measured as the similarity in publishing in the same journals.

We examine the structural patterns in the cognitive representation of social networks by systematically classifying false positives and negatives. Although existing literature on Cognitive Social Structures (CSS) has begun exploring false positives and negatives by comparing actual and perceived networks, it has not differentiated simultaneous occurrences of true and false positives and negatives on network motifs, such as reciprocity and triadic closure. Here, we propose a theoretical framework to categorize three classes of errors we call imaginary network motifs as combinations of accurately and erroneously perceived ties: (a) partially false, (b) completely false, and (c) mixed false. Using four published CSS data sets, we empirically test which imaginary network motifs are significantly more or less present in different types of perceived networks than the corresponding actual networks. Our results confirm that people not only fill in the blanks as suggested in the prior research but also conceive other imaginary structures. The findings advance our understanding of perception gaps between actual and perceived networks and have implications for designing more accurate network modeling and sampling.

An abundance of centrality indices has been proposed which capture the importance of nodes in a network based on different structural features. While there remains a persistent belief that similarities in outcomes of indices is contingent on their technical definitions, a growing body of research shows that structural features affect observed similarities more than technicalities. We conduct a series of experiments on artificial networks to trace the influence of specific structural features on the similarity of indices which confirm previous results in the literature. Our analysis on 1163 real-world networks, however, shows that little of the observations on synthetic networks convincingly carry over to empirical settings. Our findings suggest that although it seems clear that (dis)similarities among centralities depend on structural properties of the network, using correlation type analyses do not seem to be a promising approach to uncover such connections.

The exponential random graph (ERGM) model is a commonly used statistical framework for studying the determinants of tie formations from social network data. To test scientific theories under ERGMs, statistical inferential techniques are generally used based on traditional significance testing using $p$-values. This methodology has certain limitations, however, such as its inconsistent behavior when the null hypothesis is true, its inability to quantify evidence in favor of a null hypothesis, and its inability to test multiple hypotheses with competing equality and/or order constraints on the parameters of interest in a direct manner. To tackle these shortcomings, this paper presents Bayes factors and posterior probabilities for testing scientific expectations under a Bayesian framework. The methodology is implemented in the R package BFpack. The applicability of the methodology is illustrated using empirical collaboration networks and policy networks.