B1356
Title: Graph estimation in high dimensional time-series
Authors: Arkaprava Roy - University of Florida (United States) [presenting]
Abstract: Multivariate time series data are routinely collected in many application areas. Although stationarity is a useful modelling assumption for any time series data, methodological developments are limited under these assumptions for multivariate time series. Under some assumptions on the autocovariance matrices, those properties are achieved for a new class of Gaussian multivariate time series. In the proposed class, the normalized multivariate time series is assumed to be some orthogonal rotation of independent univariate latent time series. To capture the graphical dependence structure among the variables, it is also proposed to sparsely estimate the marginal precision matrix and develop related computational methodologies. An efficient Markov Chain Monte Carlo (MCMC) algorithm is developed for posterior computation. Theoretical consistency properties are also studied. Excellent performance in simulations and real data applications is shown.
