Paper detail

Second-order Approximation of Exponential Random Graph Models

Exponential random graph models (ERGMs) are flexible probability models allowing edge dependency. However, it is known that, to a first-order approximation, many ERGMs behave like Erdös-Rényi random graphs, where edges are independent. In this paper, to distinguish ERGMs from Erdös-Rényi random graphs, we consider second-order approximations of ERGMs using two-stars and triangles. We prove that the second-order approximation indeed achieves second-order accuracy in the triangle-free case. The new approximation is formally obtained by Hoeffding decomposition and rigorously justified using Stein's method.