Abstract:Time series data arising in many applications nowadays are high-dimensional. A large number of parameters describe features of these time series. We propose a novel approach to modeling a high-dimensional time series through several independent univariate time series, which are then orthogonally rotated and sparsely linearly transformed. With this approach, any specified intrinsic relations among component time series given by a graphical structure can be maintained at all time snapshots. We call the resulting process an Orthogonally-rotated Univariate Time series (OUT). Key structural properties of time series such as stationarity and causality can be easily accommodated in the OUT model. For Bayesian inference, we put suitable prior distributions on the spectral densities of the independent latent times series, the orthogonal rotation matrix, and the common precision matrix of the component times series at every time point. A likelihood is constructed using the Whittle approximation for univariate latent time series. An efficient Markov Chain Monte Carlo (MCMC) algorithm is developed for posterior computation. We study the convergence of the pseudo-posterior distribution based on the Whittle likelihood for the model's parameters upon developing a new general posterior convergence theorem for pseudo-posteriors. We find that the posterior contraction rate for independent observations essentially prevails in the OUT model under very mild conditions on the temporal dependence described in terms of the smoothness of the corresponding spectral densities. Through a simulation study, we compare the accuracy of estimating the parameters and identifying the graphical structure with other approaches. We apply the proposed methodology to analyze a dataset on different industrial components of the US gross domestic product between 2010 and 2019 and predict future observations.

Mixed orthogonality graphs for continuous‐time state space models and orthogonal projections

Mixed orthogonality graphs for continuous-time state space models and orthogonal projections

Mixed orthogonality graphs for continuous-time stationary processes

Partial correlation graphs for continuous-parameter time series

An Error-in-variable Projection to Latent Structure Framework for Monitoring Technical Systems with Orthogonal Signal Components

Graphical Models of Autoregressive Moving-Average Processes

ARMA Identification of Graphical Models.

Multivariate continuous-time autoregressive moving-average processes on cones

Stochastic modeling of stationary scalar Gaussian processes in continuous time from autocorrelation data

State Space Models for Gaussian Stochastic Processes.

Graphical models for nonstationary time series

On the Structure of State-Space Models for Discrete-Time Stochastic Vector Processes

Markov Chains and Multiple Orthogonality

Generating Different Gaussian Multivariate Processes With Identical Graphs

Dependence Estimation for High Frequency Sampled Multivariate CARMA Models

Continuously indexed graphical models

Weak equivalence of local independence graphs

Bayesian Learning of Relational Graph in Semiparametric High-dimensional Time Series

Irreducibility of nonsmooth state-space models with an application to CMA-ES

Benign Landscapes of Low-Dimensional Relaxations for Orthogonal Synchronization on General Graphs