Title: Multi-connection selection and estimation
Authors: Fengrong Wei - University of West Georgia (United States) [presenting]
Abstract: Network analysis becomes popular nowadays and subjects within a network could be complexly connected through multiple relationships. A network of $N$ nodes with $p$ types of connections is investigated and a network regression approach to investigate the heterogeneous impacts of these connections on a continuous response is proposed. The structure of each connection is taken into consideration in the proposed approach to select the important connections and estimate their corresponding effects on the response simultaneously. Simulation studies demonstrate the effectiveness of the proposed approach in both connection selection and heterogeneous effect estimation. The usefulness of the proposed approach is further illustrated in a real example. Extension studies are discussed and the corresponding implementable methodologies are provided.