Statistical Testing on Directed Graphs by Surrogate Data Generation
Researchers have developed a new framework for statistical hypothesis testing on directed graphs, extending existing methods from undirected graphs. The approach defines directed graph wide-sense stationary signals and generates surrogate signals that preserve covariance structure. This allows for the construction of null distributions to serve as a reference for empirical data, demonstrating superior performance compared to existing techniques. AI
IMPACT Introduces novel statistical methods for analyzing complex graph structures, potentially improving machine learning models that rely on relational data.