Abstract:We study estimation of large Dynamic Factor models implemented through the Expectation Maximization (EM) algorithm, jointly with the Kalman smoother. We prove that as both the cross-sectional dimension, $n$, and the sample size, $T$, diverge to infinity: (i) the estimated loadings are $\sqrt T$-consistent, asymptotically normal and equivalent to their Quasi Maximum Likelihood estimates; (ii) the estimated factors are $\sqrt n$-consistent, asymptotically normal and equivalent to their Weighted Least Squares estimates. Moreover, the estimated loadings are asymptotically as efficient as those obtained by Principal Components analysis, while the estimated factors are more efficient if the idiosyncratic covariance is sparse enough.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the problem of parameter estimation and inference in high - dimensional dynamic factor models (Dynamic Factor Models, DFM) by using the Expectation Maximization (EM) algorithm jointly with the Kalman Smoother. Specifically, the author has studied the asymptotic properties of factor loadings and factors estimated by the EM algorithm when the cross - sectional dimension $n$ and the sample size $T$ both tend to infinity. The main objectives include: 1. **Estimation of factor loadings**: - Prove that as $n$ and $T$ tend to infinity, the factor loadings estimated by the EM algorithm are consistent and asymptotically normally distributed, which is equivalent to the Quasi - Maximum Likelihood (QML) estimation. - Further prove that these estimated factor loadings are as efficient as the estimates obtained by Principal Components (PC) analysis in the asymptotic sense. 2. **Estimation of factors**: - Prove that the factors estimated by using the EM algorithm with the Kalman Smoother are also consistent and asymptotically normally distributed, which is equivalent to the Weighted Least Squares (WLS) estimation. - If the covariance matrix of the idiosyncratic components is sparse enough, then the factors estimated by the Kalman Smoother are more efficient than those by Principal Components (PC) analysis. 3. **Theoretical contributions**: - Prove that the factor loadings estimated by the EM algorithm converge to the unique maximum of the likelihood function when $n$ and $T$ tend to infinity. - Prove that the factors estimated by the Kalman Smoother are equivalent to the weighted least squares estimates when the factor loadings and the variances of the idiosyncratic components are observed when $n$ and $T$ tend to infinity. - Compare the asymptotic efficiencies of the EM algorithm and Principal Components (PC) analysis in estimating factor loadings and factors. Through these studies, the author aims to provide a solid theoretical basis for the wide application of the EM algorithm in high - dimensional dynamic factor models and respond to some long - standing criticisms of the EM algorithm, such as the view that Principal Components (PC) analysis is a better method. In addition, the author also provides the asymptotic distributions of the estimators of the EM algorithm and the Kalman Smoother, filling the theoretical gap in this field.

Quasi Maximum Likelihood Estimation and Inference of Large Approximate Dynamic Factor Models via the EM algorithm

Quasi Maximum Likelihood Estimation of High-Dimensional Factor Models: A Critical Review

Efficient Estimation of Approximate Factor Models via Regularized Maximum Likelihood

Quasi Maximum Likelihood Estimation for Large-Dimensional Matrix Factor Models

The sparse dynamic factor model: a regularised quasi-maximum likelihood approach

Inferential Theory for Factor Models of Large Dimensions

Optimal Estimation of Large-Dimensional Nonlinear Factor Models

On the Penalized Maximum Likelihood Estimation of High-Dimensional Approximate Factor Model

Simultaneous Inference For High-Dimensional Approximate Factor Model

Large-Dimensional Dynamic Factor Models: Estimation of Impulse-Response Functions with $I(1)$ Cointegrated Factors

Large Dimensional Time-Varying GMM Estimation: A New Approach

Inference for Large Dimensional Factor Models under General Missing Data Patterns

Estimation and inference for high dimensional factor model with regime switching

Approximate Factor Models with Strongly Correlated Idiosyncratic Errors

Factor modelling for high-dimensional functional time series

Likelihood approach to dynamic panel models with interactive effects

State-Varying Factor Models of Large Dimensions

On Time-Varying Factor Models: Estimation and Testing

Dynamic Sparse Factor Analysis

STATISTICAL ANALYSIS OF FACTOR MODELS OF HIGH DIMENSION

Semiparametric Conditional Factor Models: Estimation and Inference