Subsampling Based Inference for Large Networks

Large-scale networks are being generated in many scientific fields, including biological sciences, social sciences, and physical sciences. The project addresses the urgent need to develop scalable subsampling algorithms for statistical inference on such networks and provide theoretical guarantees on the performance. The proposed methods will be applied to biological and social network data, and will be used to study mindfulness-based therapies for disorders associated with hearing loss, such as tinnitus.

The project offers opportunities for involvement of graduate and undergraduate students with diverse backgrounds and interests. The proposed methods will be incorporated into relevant courses. Research results will be disseminated to the scientific communities, and all software developed in this research will be freely distributed as open-source to the public.

The project will develop subsampling based methods for inference problems, such as model selection and hypothesis testing, for large-scale networks, and investigate theoretical properties of these methods to provide statistical guarantees on performance. The subsampling strategies will be applied to a broad range of models for networks, including stochastic block models, random dot product graph models, latent space models, and other models for networks.

The theoretical properties of subsampling methods investigated in this project include the consistency of model selection, hypothesis testing, and parameter estimation. The proposed subsampling methods will be applied to real network data from social and natural sciences.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.