Premium
Topology and Dependency Tests in Spatial and Network Autoregressive Models
Author(s) -
Farber Steven,
Páez Antonio,
Volz Erik
Publication year - 2009
Publication title -
geographical analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.773
H-Index - 65
eISSN - 1538-4632
pISSN - 0016-7363
DOI - 10.1111/j.1538-4632.2009.00751.x
Subject(s) - dependency (uml) , network topology , computer science , hierarchical network model , topology (electrical circuits) , autoregressive model , hierarchy , context (archaeology) , cluster analysis , spatial analysis , data mining , econometrics , mathematics , artificial intelligence , statistics , geography , archaeology , combinatorics , economics , market economy , operating system
Social network analysis has been identified as a promising direction for further applications of spatial statistical and econometric models. The type of network analysis envisioned is formally identical to the analysis of geographical systems, in that both involve the measurement of dependence between observations connected by edges that constitute a system. An important item, which has not been investigated in this context, is the potential relationship between the topology properties of networks (or network descriptions of geographical systems) and the properties of spatial models and tests. The objective of this article is to investigate, within a simulation setting, the ability of spatial dependency tests to identify a spatial/network autoregressive model when two network topology measures, namely degree distribution and clustering, are controlled. Drawing on a large data set of synthetically controlled social networks, the impact of network topology on dependency tests is investigated under a hierarchy of topology factors, sample size, and autocorrelation strength. In addition, topology factors are related to known properties of empirical systems.