12.07.2015 Views

Sunbelt XXXI International Network for Social Network ... - INSNA

Sunbelt XXXI International Network for Social Network ... - INSNA

Sunbelt XXXI International Network for Social Network ... - INSNA

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Generosity As A Public Good: Heterogeneous Preferences In Partner Selection Promote Opinion Diversity And <strong>Social</strong> IntegrationSohn, Yunkyu<strong>Network</strong> DynamicsPeer Influence, Co‐Evolution ModelTHURS.PM2Although polarization of public opinion and segregation of social groups are prevalent, we rarely observe complete fragmentation of a social network. Toexplain this regularity, we propose a utility‐based model of opinion <strong>for</strong>mation and partner selection. The model allows feedback dynamics between opinionsand social relationships based on individual tolerance thresholds of opinion difference. Our model differs notably from the existing co‐evolutionary models inthe following aspects: i) Each dyadic relationship is weighted asymmetrically; ii) Opinion change is a Markov process; iii) In an equilibrium phase, where thedifference between opinions of agents holding a directed arc is lower than the tolerance threshold of the receiver, the resulting network and agents can havevarious topologies and opinion distributions. We investigate topological characteristics and opinion distribution of the model at equilibria when a) ahomogenous tolerance threshold is allowed and b) heterogeneous thresholds are distributed over the population. In contrast to the homogeneous thresholdcase where networks at equilibria exhibit either complete fragmentation by opinion or <strong>for</strong>m fully connected components holding a homogenous opinion, theheterogeneous setting simultaneously achieves opinion diversity and social integration by binding structurally modularized opinion clusters with each other.The topological properties of these networks surprisingly resemble those of real social networks.Goodness Of Fit For <strong>Social</strong> <strong>Network</strong> DynamicsLospinoso, Joshua A.; Snijders, Tom A.Analyzing <strong>Network</strong> DataStatistical Methods, Dynamic <strong>Network</strong> Analysis, Siena, Goodness‐of‐fit, Actor‐based Models, DegeneracySAT.PM1We propose new statistical procedures <strong>for</strong> evaluating the goodness of fit of stochastic actor oriented models (SAOMs) <strong>for</strong> social network dynamics. Due to theunique nature of longitudinal social network data (a single observed trajectory), classic tests <strong>for</strong> goodness of fit are generally inappropriate. We develop (1) ageneral non‐parametric test based on auxiliary features of the network like triad census counts, behavioral profiles, geodesic distances, etc., which leverages asimulated cumulative density function (2) an in<strong>for</strong>mation criteria <strong>for</strong> comparing among likelihood based models <strong>for</strong> social network dynamics based on pathintegration, and (3) a series of parametric tests <strong>for</strong> model selection among nested SAOMs based on the classic asymptotic tests <strong>for</strong> composite hypotheses. Wepresent a simulation studies to illustrate the effectiveness of these approaches, and apply them as a systematic examination to a real world dataset.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!