A Pure Hypothesis Test for Inhomogeneous Random Graph Models Based on a Kernelised Stein Discrepancy
This addresses the need for reliable statistical tests for network models, especially for small or non-asymptotic networks, which is incremental as it adapts an existing KSD method to a specific graph context.
The authors tackled the problem of goodness-of-fit testing for inhomogeneous random graph models in high-dimensional settings, developing a kernelised Stein discrepancy test that works with a single network observation and applies to networks of any size, particularly small ones where asymptotic tests fail, with theoretical guarantees provided.
Complex data are often represented as a graph, which in turn can often be viewed as a realisation of a random graph, such as an inhomogeneous random graph model (IRG). For general fast goodness-of-fit tests in high dimensions, kernelised Stein discrepancy (KSD) tests are a powerful tool. Here, we develop a KSD-type test for IRG models that can be carried out with a single observation of the network. The test applies to a network of any size, but is particularly interesting for small networks for which asymptotic tests are not warranted. We also provide theoretical guarantees.