Abstract:This article proposes a new approach to modeling high-dimensional time series by treating a $p$-dimensional time series as a nonsingular linear transformation of certain common factors and idiosyncratic components. Unlike the approximate factor models, we assume that the factors capture all the non-trivial dynamics of the data, but the cross-sectional dependence may be explained by both the factors and the idiosyncratic components. Under the proposed model, (a) the factor process is dynamically dependent and the idiosyncratic component is a white noise process, and (b) the largest eigenvalues of the covariance matrix of the idiosyncratic components may diverge to infinity as the dimension $p$ increases. We propose a white noise testing procedure for high-dimensional time series to determine the number of white noise components and, hence, the number of common factors, and introduce a projected Principal Component Analysis (PCA) to eliminate the diverging effect of the idiosyncratic noises. Asymptotic properties of the proposed method are established for both fixed $p$ and diverging $p$ as the sample size $n$ increases to infinity. We use both simulated data and real examples to assess the performance of the proposed method. We also compare our method with two commonly used methods in the literature concerning the forecastability of the extracted factors and find that the proposed approach not only provides interpretable results, but also performs well in out-of-sample forecasting. Supplementary materials of the article are available online.

Bayesian Estimation of Categorical Dynamic Factor Models

Bayesian Dynamic Factor Models for High-dimensional Matrix-valued Time Series

Bayesian factor models for multivariate categorical data obtained from questionnaires

A Bayesian Dynamic Factor-Residual Model for Spatial Dependence in Longitudinal Data

Dynamic Factor Models for Multivariate Count Data: An Application to Stock-Market Trading Activity

Bayesian Dynamic Mediation Analysis

Analyzing Multiple Multivariate Time Series Data Using Multilevel Dynamic Factor Models

Bayesian mixtures of common factor analyzers: Model, variational inference, and applications

Estimation of dynamic mixed double factors model in high-dimensional panel data

Bayesian Statistical Inference for Factor Analysis Models with Clustered Data

Dynamic Sparse Factor Analysis

Comparisons of Four Methods for Estimating a Dynamic Factor Model

Path and Direction Discovery in Individual Dynamic Factor Models: A Regularized Hybrid Unified Structural Equation Modeling with Latent Variable

Dynamic partial (co)variance forecasting model

Factor-augmented Bayesian treatment effects models for panel outcomes

Dynamic Matrix Factor Models for High Dimensional Time Series

Modeling High-Dimensional Time Series: A Factor Model with Dynamically Dependent Factors and Diverging Eigenvalues

FactorVAE: A Probabilistic Dynamic Factor Model Based on Variational Autoencoder for Predicting Cross-Sectional Stock Returns

State-Varying Factor Models of Large Dimensions

Dynamic Factor Analysis of High-dimensional Recurrent Events