The Dynamic, the Static, and the Weak: Factor models and the analysis of high-dimensional time series

2024-07-15

AI Summary

This paper addresses six core controversies in high-dimensional time series factor modeling: dynamic versus static loadings, strong versus weak factors, weak common components (Gersing et al., 2023), cross-sectional rank invariance, unidentifiable strong factor interference, and integration of common and idiosyncratic forecasts. Building upon Forni et al.’s (2000) Generalized Dynamic Factor Model (GDFM) framework, we develop a unified diagnostic methodology that integrates frequency-domain analysis, asymptotic principal component estimation, and asymptotic theory of factor strength to reliably identify weak common factors and latent strong factors. We establish theoretical guarantees showing that the GDFM achieves superior factor identification rates, prediction consistency, and model robustness compared to static approximations. The results provide a more rigorous theoretical foundation and an operationally feasible empirical paradigm for high-dimensional time series modeling.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems
• Machine Learning: Other Foundations of Machine Learning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Several fundamental and closely interconnected issues related to factor models are reviewed and discussed: dynamic versus static loadings, rate-strong versus rate-weak factors, the concept of weakly common component recently introduced by Gersing et al. (2023), the irrelevance of cross-sectional ordering and the assumption of cross-sectional exchangeability, the impact of undetected strong factors, and the problem of combining common and idiosyncratic forecasts. Conclusions all point to the advantages of the General Dynamic Factor Model approach of Forni et al. (2000) over the widely used Static Approximate Factor Model introduced by Chamberlain and Rothschild (1983).

Problem

• Comparing dynamic vs static factor models in time series
• Analyzing rate-strong vs rate-weak factors in high dimensions
• Evaluating impact of undetected strong factors on forecasts

Innovation

• Dynamic loadings in factor models
• Weakly common component concept
• General Dynamic Factor Model advantages