Abstract:
|
For multivariate functional data with plentiful functional components, an efficient dimension reduction approach to capture features is desired. We propose a functional latent factor model, where complex covariance is decomposed into a shared term and an outcome-specific term. This model uses low dimensional latent factors along with coefficient vectors to automatically induce dependency among observed functions. For data with higher dimensions, a sparse structure in coefficient vectors is further considered. We illustrate the performance of the proposed model through simulation studies and an application to electroencephalogram(EEG) data obtained by 64 electrodes.
|