Not logged in.
Quick Search - Contribution
Contribution Details
Type | Conference Presentation |
Scope | Discipline-based scholarship |
Title | The effects the number of agents has in the formation of networks and statistical analysis on multiple networks |
Organization Unit | |
Authors |
|
Presentation Type | speech |
Item Subtype | Original Work |
Refereed | Yes |
Status | Published in final form |
Language |
|
Event Title | INSNA 2016 Sunbelt conference |
Event Type | conference |
Event Location | Newport beach, California, USA |
Event Start Date | April 5 - 2016 |
Event End Date | April 10 - 2016 |
Date Annual Report | 2016 |
Abstract Text | Which are the mechanisms that underlie the formation of networks? This is a key question in network science, pervading the most variegated disciplines, and extensively ap- proached both, theoretically and empirically. Strikingly, most results depend on assuming a large number of network constituents and little is known for networks in which this is implausible. For instance, most existing statistical network analyses rely on large sample properties of estimators, where sample is defined on the network size. However, as we show in this Paper, several statistical network studies based on observational data suffer from two shortcomings: (1) they are not replicable, since parameters are not constant on sample size – as opposed with other regression models- and (2) for the exponential random graph model (ERGM) – among the most used statistical network models – large sample properties remain unknown. Here, we address the first problem by determining the functional form of the parameters on the number of agents for given ERGMs. We use these results to con- struct a class of models termed finite exponential random graph model (fERGM), which do not make assumptions on the network size, but on the number of observed networks. This exchange of assumptions proves fundamental for the study of the influence the network size on the network formation. We also demonstrate that a recent methodology for addressing environment effects (e.g. like the network size) has on the formation of network lacks on fundamental statistical properties, and thus some empirical results need to be revised. Finally, we demonstrate how to use fERGM to test for the effect that the network sizes has on the simple mechanisms for the formation of networks, i.e. reciprocity and transitivity. |
Export | BibTeX |
Funders | URPP Social Network |