FNETS: Factor-Adjusted Network Estimation and Forecasting for High-Dimensional Time Series

Barigozzi, Matteo; Cho, Haeran; Owens, Dom

doi:10.1080/07350015.2023.2257270

Cited by 10 publications

(11 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, we introduce the factor-adjusted VAR model and describe the FNETS methodology proposed in Barigozzi et al (2023) for network estimation and forecasting of high-dimensional time series. We limit ourselves to describing the key steps of FNETS and refer to the above paper for its comprehensive treatment.…”

Section: Fnets Methodologymentioning

confidence: 99%

“…Throughout, we refer to the models ( 2) and ( 3) as unrestricted and restricted to highlight that the latter imposes more restrictions on the model. Barigozzi et al (2023) propose a factor-adjusted VAR model under which we observe a zero-mean, second-order stationary process X t = (X 1t , . .…”

Section: Factor-adjusted Var Modelmentioning

confidence: 99%

“…Additionally, the left panel of Figure 2 shows the inadequacy of fitting a VAR model to such data under the sparsity assumption via ℓ 1 -regularisation methods, unless the presence of strong correlations is accounted for by a factor-adjustment step as in the right panel. Barigozzi et al (2023) propose the FNETS method for factor-adjusted VAR modelling of highdimensional, second-order stationary time series. Under their proposed model, the data is decomposed into two latent components such that the factor-driven component accounts for pervasive leading, lagging or contemporaneous co-movements of the variables, while the remaining idiosyncratic dynamic dependence between the variables is modelled by a sparse VAR process.…”

Section: Introductionmentioning

confidence: 99%

“…

…”

mentioning

confidence: 99%

See 3 more Smart Citations

fnets: An R Package for Network Estimation and Forecasting via Factor-Adjusted VAR Modelling

Owens,

Cho,

Barigozzi

2023

The R Journal

View full text Add to dashboard Cite

Vector autoregressive (VAR) models are useful for modelling high-dimensional time series data. This paper introduces the package fnets, which implements the suite of methodologies proposed by (Barigozzi, Cho, and Owens 2023) for the network estimation and forecasting of high-dimensional time series under a factor-adjusted vector autoregressive model, which permits strong spatial and temporal correlations in the data. Additionally, we provide tools for visualising the networks underlying the time series data after adjusting for the presence of factors. The package also offers data-driven methods for selecting tuning parameters including the number of factors, the order of autoregression, and thresholds for estimating the edge sets of the networks of interest in time series analysis. We demonstrate various features of fnets on simulated datasets as well as real data on electricity prices.

show abstract

Section: Fnets Methodologymentioning

confidence: 99%

Section: Factor-adjusted Var Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…

…”

mentioning

confidence: 99%

See 2 more Smart Citations

fnets: An R Package for Network Estimation and Forecasting via Factor-Adjusted VAR Modelling

Owens,

Cho,

Barigozzi

2023

The R Journal

View full text Add to dashboard Cite

show abstract

“…Omitted relevant factors from the financial markets can lead to misleading results. Some researchers (Kock and Callot, 2015) tried to handle this shortcoming of the original approach with a FAVAR model estimation ( (Claeys and Vašíček, 2014;Barigozzi et al, 2021Barigozzi et al, , 2023, but most articles do not address this issue.…”

Section: Limitations Of the Dy Frameworkmentioning

confidence: 99%

Network Analysis of the Financial Sector: A Comprehensive Perspective with Adaptive Joint LASSO Method

Badics

View full text Add to dashboard Cite

The emergence of systemic risk from the 2008 Global Financial Crisis (GFC) and the recent energy market shocks following the COVID-19 pandemic and the onset of the Russo-Ukrainian war highlighted the importance of understanding shock spillovers for policymakers and academics alike. Specifically, shock transmissions in financial network settings warrant attention to support regulatory or policy interventions for effectively mitigating or preventing the transmission of systemic risk. The thesis focuses on this important topic from various angles to answer the following questions: • How can machine learning techniques help to improve financial network studies, the analysis of high dimensional time series? • How can structural changes in financial networks efficiently examined using event analysis framework? My thesis provides methodological and conceptual contributions to financial network analysis. The contribution of my thesis to the existing literature is threefold: 1. I propose a new regularization method, the adaptive joint least absolute shrinkage and selection operator (AJ LASSO). The innovation in this method is that it accounts for possible sparsity both in the coefficient and the covariance matrix of the estimated Vector autoregressive (VAR) model. The method is especially suitable for high-dimension network analysis, where parameter estimation is a critical issue. 2. I extend the Diebold-Yilmaz (DY) framework with an event study tool to facilitate an in-depth analysis of the contagion channels during critical local and global shocks in financial networks. Integrating a moving-block bootstrap method (MBB) into the framework, I can investigate how the observed shocks transform the financial networks. 3. My conceptual contribution is that I characterize and analyze illiquidity networks. The liquidity concerns are globally recognized and the focal point of the newly revised Basel IV regulatory guidelines; thus, intuitively and from a regulatory perspective, illiquidity spillover analysis in the financial sector is valuable. I analyze the illiquidity connectedness of financial institutions (FIs) and show that the illiquidity network better tracks the dominant shock transmitters in the system during financial turmoil than the volatility-based networks. I show that the DY framework extended with the MBB method is a powerful tool to identify troubled financial institutions and contagion channels.

show abstract

Testing for sparse idiosyncratic components in factor-augmented regression models

Beyhum,

Striaukas

2024

Journal of Econometrics

View full text Add to dashboard Cite

FNETS: Factor-Adjusted Network Estimation and Forecasting for High-Dimensional Time Series

Cited by 10 publications

References 50 publications

fnets: An R Package for Network Estimation and Forecasting via Factor-Adjusted VAR Modelling

fnets: An R Package for Network Estimation and Forecasting via Factor-Adjusted VAR Modelling

Network Analysis of the Financial Sector: A Comprehensive Perspective with Adaptive Joint LASSO Method

Testing for sparse idiosyncratic components in factor-augmented regression models

Contact Info

Product

Resources

About