Exact tests for stochastic block models

We develop a finite-sample goodness-of-fit test for \emph{latent-variable} block models for networks and test it on simulated and real data sets. The main building block for the latent block assignment model test is the exact test for the model with observed blocks assignment. The latter is implemented using algebraic statistics. While we focus on three variants of the stochastic block model, the methodology extends to any mixture of log-linear models on discrete data.