The method of moments and degree distributions for network models

Abstract

Probability models on graphs are becoming increasingly important in many applications, but statistical tools for fitting such models are not yet well developed. Here we propose a general method of moments approach that can be used to fit a large class of probability models through empirical counts of certain patterns in a graph. We establish some general asymptotic properties of empirical graph moments and prove consistency of the estimates as the graph size grows for all ranges of the average degree including $\Omega(1)$. Additional results are obtained for the important special case of degree distributions.Comment: Published in at http://dx.doi.org/10.1214/11-AOS904 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org

Consistent Latent Position Estimation and Vertex Classification for Random Dot Product Graphs

“…Now, we will prove (4). Note that because the fX i g are i.i.d., the f b X i g are exchangeable and hence identically distributed.…”

“…½0; 1. Bickel and Chen [3] provide a concise overview of the details of this framework, and Bickel et al [4] consider estimation for general exchangeable random graphs using the method of moments.…”

Review of statistical network analysis: models, algorithms, and software

“…Bickel and Chen [115] studied the connections between modularity (Section 2.4) and stochastic block models, while Bickel et al . [116] discuss some of the asymptotic properties of a class of models of which stochastic blockmodels are a subset. Rohe et al .…”

A Network Structural Approach to the Link Prediction Problem

“…Unfortunately, it is often difficult to obtain accurate estimation by such methods when the network is evolving. Instead of fitting directly to any distribution functions, it is possible to use hidden moments of networks, which eventually determine how the networks evolve over time (Bickel et al 2011). Incorporating the moments for degree distribution into the link prediction approach instead of the degree distribution might enable us to avoid difficulties in estimating the degree distribution precisely for the evolving networks.…”