EcoSta 2023: Start Registration
View Submission - EcoSta2023
A0895
Title: A general framework for network autoregressive processes Authors:  George Michailidis - University of California, Los Angeles (United States) [presenting]
Abstract: A general, flexible framework for Network Autoregressive Processes (NAR) is developed, wherein the response of each node in the network linearly depends on its past values, a prespecified linear combination of neighbouring nodes and a set of node-specific covariates. The corresponding coefficients are node-specific, and the framework can accommodate heavier than Gaussian errors with spatial-autoregressive, factor-based, or in specific settings, general covariance structures. A sufficient condition is provided that ensures the stability (stationarity) of the underlying NAR that is significantly weaker than its counterparts in previous work in the literature. Further, ordinary and (estimated) generalized least squares estimators are developed for both fixed and diverging numbers of network nodes and provide their ridge regularized counterparts that exhibit better performance in large network settings, together with their asymptotic distributions. Their asymptotic distributions are derived that can be used for testing various hypotheses of interest to practitioners. The issue of misspecifying the network connectivity and its impact on the aforementioned asymptotic distributions of the different NAR parameter estimators are also addressed. The framework is illustrated on both synthetic and real air pollution data.