A1378
Title: Mean independent component analysis for multivariate time series
Authors: Chung Eun Lee - Baruch College (United States) [presenting]
Zeda Li - Baruch College (United States)
Abstract: Mean independent component analysis for multivariate time series is introduced to reduce the parameter space. In particular, a contemporaneous linear transformation is sought that detects univariate mean independent components so that each component can be modeled separately. The mean independent component analysis is flexible in the sense that no parametric model or distributional assumptions are made. A unified framework is proposed to estimate the mean independent components from data with fixed dimension or diverging dimension. The mean independent components are estimated by martingale difference divergence so that mean dependence across components and across time is minimized. The approach is extended to group mean independent component analysis by imposing a group structure on the mean independent components. A method is further introduced to identify the group structure when it is unknown. The consistency of both proposed methods is established. Extensive simulations and a real data illustration for community mobility are provided to demonstrate the efficacy of the method.
