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

Measurement Error In Network Analysis: The Effects Of Missing, Spurious, And Aggregated DataWang, Dan J.; Shi, Xiaolin; McFarland, Daniel A.; Leskovec, JureAnalyzing Network Data / BlockmodelingMethods, Centrality, Missing DataSAT.AM2We embed missing data in a broader classification of measurement error scenarios. In addition to missing data, which we term "false negative nodes andedges", we analyze the consequences of "false positive nodes and edges", and "falsely aggregated and disaggregated nodes". We simulate these sixmeasurement errors in an online social network and a publication citation network, reporting their effects on four node‐level measures‐‐degree centrality,clustering coefficient, network constraint, and eigenvector centrality. Our results suggest that in networks with more positively‐skewed degree distributionsand higher average clustering, these measures tend to be less resistant to most forms of measurement error. In addition, we argue that the sensitivity of agiven measure to an error scenario depends on the nature of the measure's calculation. Thus, we revise the claim that the more `global' a measure, the lessresistant it is to measurement error. We find that clustering coefficient and network constraint are, in general, the least resistant of our four measures eventhough they are less `global' measures than eigenvector centrality. Finally, we anchor our discussion to examples of past network research that likely sufferfrom these different forms of measurement error while making recommendations for error correction strategies.Meeting In Settings, Mating In Networks: Stochastic Actor‐oriented Models For Large Network DynamicsPreciado, Paulina; Snijders, Tom A.; Lospinoso, Joshua A.Network DynamicsIncomplete information, Settings, Meeting and mating, Stochastic Actor‐ Oriented Models, Large network dynamicsTHURS.PM1Stochastic Actor‐Oriented Models (SAOM; Snijders, 2001), provide a flexible and rich framework for analysing social network dynamics. These models assumethat actors have full information about the network and about the other actors. For large networks the assumption of full information is often not realistic. Inthis paper we extend the SAOM by proposing that social actors are embedded in settings (Feld, 1981; Pattison & Robins, 2002) in which they have differentlevels of information about the network and the other actors. In their primary setting, which includes their direct network neighbourhood, actors have fullinformation about the other actors and the ties between them, and the usual SAOM specification operates. Besides, actors can have meeting settings in whicha two‐step process takes place: at a given moment, a particular actor i randomly meets another actor j (“meeting”), and acquires information thatprobabilistically determines the creation of a tie (mating). Examples of meeting settings are institutions (schools, work places), neighbourhoods and theresidual category of the rest of the network. We compare the standard SAOM, with the extended settings specification using a three‐wave friendship networkof all adolescents in an age cohort living in a small Swedish town. Computing time for large networks is reduced by the new modelling framework.