Abstract:
|
Multiple-network data has attracted increasing attention recently, where the data are recorded as symmetric matrices, and each matrix encodes an individual network structure. Such data arises frequently in various scientific fields such as the analyses of brain connectivity and gene interactions. In these studies, it is of great interest to compare the means of two populations of networks. In this work, we propose a hypothesis testing procedure when we are interested in a given area of networks that may have multiple signals functioning together. We establish asymptotic results for the proposed test under general moment conditions and validate the test under a variety of popular network models. We further demonstrate the efficacy of the proposed test under simulation studies and the analysis of a brain dataset.
|