B1984
Title: Adaptive distributed inference for multi-source massive heterogeneous data
Authors: Mixia Wu - Beijing University of Technology (China) [presenting]
Abstract: Distributed inference for a heterogeneous linear model with massive datasets is considered. The goal is to extract common features across all subpopulations and explore the heterogeneity of each subpopulation. Noticing that heterogeneity may exist not only in expectations of subpopulations but also in their variances, we propose the heterogeneity-adaptive distributed aggregation (HADA) estimation, which is shown to be communication-efficient and asymptotically optimal, irrespective of homoscedasticity or heteroscedasticity. Furthermore, a distributed test for parameter heterogeneity across sub-populations is constructed based on the HADA estimator. The finite-sample performance of the proposed methods is evaluated via simulation studies and bike-share data.
