A0503
Title: Shrinkage estimation and inference for network-linked data: A robust spectral embedding approach
Authors: Feng Zhang - Xiamen University (China) [presenting]
Chen Yahui - Xiamen University (China)
Han Xiaoyi - Xiamen University (China)
Abstract: We present a regression framework that integrates spectral network information to enhance the efficiency and robustness of modeling network-linked data. We utilize the broken adaptive ridge shrinkage estimator to identify the sparsity structure of network-irrelevant covariates, complemented by a subspace projection method for effect separation. We introduce a robust covariance estimator tailored for heteroskedastic disturbances. We rigorously establish the theoretical properties of our estimator and demonstrate its efficacy through comprehensive simulation experiments across diverse scenarios. For the application, we analyze a statistical paper citation network. In this study, we not only detect significant network effects but also provide interpretable insights with the proposed method. This work contributes to the growing literature on network data modeling by offering a statistically sound, robust, and interpretable framework for analyzing complex network-linked data.
