Estimating Inhomogeneous Spatial Interaction Functions for Spatial Network Simulation
Loring J Thomas, Peng Huang, Iris Luo, John Hipp, Carter T ButtsPhysical Distance has a variety of effects on the structure of social networks. As distance increases, the probability of a given network tie decreases. While prior research has often examined this relationship assuming homogeneous populations, it is likely the case that the form of this effect is different for different groups, primarily due to the lack of population data that has high spatial resolution. Using data from the American Social Fabric project and small scale areal unit estimation of joint gender/race/ethnicity distributions within the United States, we estimate a series of inhomogeneous Spatial Interaction Functions (SIFs) that are dependent on a combination of demographic features. From these SIFs, we describe a class of Bernoulli graphs that are conditioned on distance, and present results pertaining to the form of the SIF and simulated graph structures for different groups.