pith. machine review for the scientific record. sign in

arxiv: 2604.12563 · v1 · submitted 2026-04-14 · 📊 stat.ME · econ.EM

Recognition: unknown

Latent community paths in VAR-type models via dynamic directed spectral co-clustering

Younghoon Kim , Changryong Baek
Authors on Pith no claims yet

Pith reviewed 2026-05-10 14:57 UTC · model grok-4.3

classification 📊 stat.ME econ.EM
keywords VAR modelsdynamic networksspectral co-clusteringcommunity detectiondirected graphsstochastic blockmodeltime serieshigh-dimensional statistics
0
0 comments X

The pith

Embedding a degree-corrected stochastic co-blockmodel into VAR transition matrices lets directed spectral co-clustering recover evolving sending and receiving communities with explicit error bounds.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a framework that treats the transition matrices of high-dimensional VAR-type processes as realizations of a degree-corrected stochastic co-blockmodel whose community structure can change over time. By applying directed spectral co-clustering followed by eigenvector smoothing, the method extracts low-dimensional summaries that distinguish nodes' outgoing and incoming roles and tracks how those groups split, merge, or persist. It supplies non-asymptotic misclassification bounds that hold for both periodic VAR models with seasonal cycles and generalized VHAR models with ordered dependence horizons. The approach is illustrated on U.S. payroll employment series and global equity volatility data, where it identifies stable core-periphery patterns alongside more mobile sectoral re-allocations.

Core claim

The central claim is that embedding a degree-corrected stochastic co-blockmodel into the transition matrices of VAR-type systems separates sending and receiving roles at the node level, reduces complex directional dependence to an interpretable low-dimensional form, and permits consistent recovery of latent community paths via directed spectral co-clustering plus eigenvector smoothing, with non-asymptotic misclassification bounds holding for both periodic VAR and generalized VHAR specifications.

What carries the argument

Directed spectral co-clustering with eigenvector smoothing applied to transition matrices that arise from a degree-corrected stochastic co-blockmodel.

If this is right

  • The same embedding and smoothing procedure yields consistent community recovery for both cyclical PVAR models and multi-horizon VHAR models.
  • Node-level separation of sending and receiving roles produces an interpretable low-dimensional representation of directional dependence.
  • Non-asymptotic misclassification bounds quantify the finite-sample reliability of the recovered paths.
  • Applications to U.S. nonfarm payrolls and global stock volatilities distinguish stable core blocks from seasonally mobile or re-allocating sectors.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The framework could be applied to other linear multivariate time-series specifications whose coefficient matrices admit a low-rank block approximation.
  • The recovered sending and receiving labels might serve as time-varying covariates in subsequent regression or policy models.
  • Relaxing the smoothness assumption on community evolution would require alternative smoothing operators or change-point detection steps.
  • The method's performance on networks with heavy-tailed degree distributions could be checked by varying the degree-correction parameters in simulation.

Load-bearing premise

The transition matrices of the observed VAR-type process are generated from or well approximated by a degree-corrected stochastic co-blockmodel whose community labels change smoothly enough for eigenvector smoothing to recover them.

What would settle it

A controlled simulation in which community labels are forced to jump discontinuously or the transition matrices are generated from a non-block structure, after which the misclassification rate of the recovered paths exceeds the derived bound.

Figures

Figures reproduced from arXiv: 2604.12563 by Changryong Baek, Younghoon Kim.

Figure 1
Figure 1. Figure 1: Community paths used in the ScBM-PVAR simulation. Path 1 is static, Path 2 follows [PITH_FULL_IMAGE:figures/full_fig_p018_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Community paths used in the generalized ScBM-VHAR simulation. Path 1 is static, [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Aligned quarterly community paths of the 22 industry sectors in U.S. nonfarm payroll [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Sankey diagram of horizon-aggregated community memberships for realized volatilities [PITH_FULL_IMAGE:figures/full_fig_p024_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: ARI comparison for the ScBM-PVAR models (average of type1 and type2). [PITH_FULL_IMAGE:figures/full_fig_p042_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: ARI comparison for the generalized ScBM-VHAR models (average of type1 and type2). [PITH_FULL_IMAGE:figures/full_fig_p043_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: (Top) Time plots of the first eight time series (see Table 5 for code). (Left Bottom) Sample [PITH_FULL_IMAGE:figures/full_fig_p045_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Scree plots of singular values from the estimated transition matrices of the ScBM-PVAR [PITH_FULL_IMAGE:figures/full_fig_p046_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Alternative aligned quarterly community paths of the 22 industry sectors in U.S. nonfarm [PITH_FULL_IMAGE:figures/full_fig_p046_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: (Top) Time plots of four selected stock indices (FTSE 100, Nikkei 225, KOSPI Composite [PITH_FULL_IMAGE:figures/full_fig_p048_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Scree plots for the realized-volatility application from the lasso estimation of the gen [PITH_FULL_IMAGE:figures/full_fig_p049_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Relative to the parsimonious 3–3–3 specification discussed in the text, the 3–4–3 setting [PITH_FULL_IMAGE:figures/full_fig_p049_12.png] view at source ↗
read the original abstract

This paper proposes a dynamic network framework for uncovering latent community paths in high-dimensional VAR-type models. By embedding a degree-corrected stochastic co-blockmodel (ScBM) into the transition matrices of VAR-type systems, we separate sending and receiving roles at the node level and summarize complex directional dependence in an interpretable low-dimensional form. Our method integrates directed spectral co-clustering with eigenvector smoothing to track how directional groups split, merge, or persist over time. This framework accommodates both periodic VAR (PVAR) models for cyclical seasonal evolution and generalized VHAR models for structural transitions across ordered dependence horizons. We establish non-asymptotic misclassification bounds for both procedures and provide supporting evidence through Monte Carlo experiments. Applications to U.S.\ nonfarm payrolls distinguish a recurrent business-centered core from more mobile, seasonally sensitive sectors. In global stock volatilities, the results reveal a compact U.S.-centered long-horizon block, a Europe-heavy developed core, and a more dynamic short-horizon reallocation of peripheral and bridge markets.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript proposes a dynamic network framework for uncovering latent community paths in high-dimensional VAR-type models. By embedding a degree-corrected stochastic co-blockmodel (ScBM) into the transition matrices, it separates sending and receiving roles at the node level via directed spectral co-clustering combined with eigenvector smoothing. The approach handles periodic VAR (PVAR) and generalized VHAR models, establishes non-asymptotic misclassification bounds for the procedures, and provides Monte Carlo experiments along with applications to U.S. nonfarm payrolls and global stock volatilities.

Significance. If the central claims hold, the work provides an interpretable low-dimensional summary of complex directional dependencies in multivariate time series while separating node roles and tracking dynamic community evolution. The non-asymptotic bounds and Monte Carlo support could advance network analysis in econometrics and statistics, particularly for time-varying structures.

major comments (1)
  1. Abstract and theoretical results on misclassification bounds: The non-asymptotic misclassification bounds are stated for the spectral co-clustering and smoothing steps under the assumption that the transition matrices are known and close to the degree-corrected ScBM. However, these matrices must be estimated from finite-length time series in the VAR-type model, and the paper does not incorporate the additional finite-sample estimation error (e.g., from least-squares or regularized estimators) into the perturbation analysis or bounds. This is load-bearing for the central claim, as the eigenvector perturbation guarantees may fail to control the final misclassification rate when VAR estimation perturbations exceed the assumed separation.
minor comments (2)
  1. Monte Carlo experiments: The description lacks explicit details on data generation, exact exclusion rules for time series, hyperparameter selection (including number of communities), and code availability, which would aid reproducibility.
  2. Notation and implementation: The choice of the number of communities (a free parameter) and the precise smoothing procedure for eigenvector paths across time could be clarified with pseudocode or additional equations.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful and constructive review of our manuscript. We address the major comment below.

read point-by-point responses

  1. Referee: The non-asymptotic misclassification bounds are stated for the spectral co-clustering and smoothing steps under the assumption that the transition matrices are known and close to the degree-corrected ScBM. However, these matrices must be estimated from finite-length time series in the VAR-type model, and the paper does not incorporate the additional finite-sample estimation error (e.g., from least-squares or regularized estimators) into the perturbation analysis or bounds. This is load-bearing for the central claim, as the eigenvector perturbation guarantees may fail to control the final misclassification rate when VAR estimation perturbations exceed the assumed separation.

    Authors: We thank the referee for highlighting this important point. The non-asymptotic misclassification bounds (Theorems 3.1 and 3.2) are indeed derived under the assumption that the transition matrices are observed and lie close to a degree-corrected stochastic co-blockmodel with sufficient separation; the perturbation analysis for the directed spectral co-clustering and eigenvector smoothing steps is conditional on this. The manuscript does not fold the finite-sample estimation error arising from least-squares or regularized estimation of the VAR-type coefficients into these bounds. This is a genuine limitation of the current theoretical results. We will revise the paper to (i) explicitly state the conditional nature of the bounds in the theoretical section, (ii) add a remark outlining how existing high-dimensional consistency results for VAR estimators can be combined with the existing perturbation bounds via a triangle inequality argument on the eigenvector error (provided the estimation error is of smaller order than the separation), and (iii) clarify that the Monte Carlo experiments already use estimated transition matrices and demonstrate practical robustness. This partial revision will make the scope of the guarantees transparent without requiring a complete re-derivation of the non-asymptotic bounds. revision: partial

Circularity Check

0 steps flagged

No significant circularity; bounds derived independently under stated ScBM assumption

full rationale

The paper embeds a degree-corrected stochastic co-blockmodel into VAR transition matrices and derives non-asymptotic misclassification bounds for the directed spectral co-clustering and eigenvector smoothing steps. These bounds are established conditionally on the transition matrices being close to the ScBM structure, without reducing the community path recovery to a fitted parameter by construction or relying on load-bearing self-citations for the core uniqueness or separation arguments. The VAR estimation step is treated as a separate preprocessing stage whose error is not folded into the clustering bounds, but this is an assumption limitation rather than a definitional loop. The derivation chain remains self-contained against external spectral clustering theory and blockmodel results.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central modeling step assumes that VAR transition matrices admit a degree-corrected stochastic co-blockmodel representation; the number of communities and any smoothing bandwidth are likely chosen or estimated but are not quantified in the abstract. The latent community paths themselves are the primary new construct without external falsifiable evidence beyond model fit.

free parameters (1)
  • number of communities
    Standard hyper-parameter in blockmodel-based clustering; value not stated in abstract.
axioms (1)
  • domain assumption Transition matrices of the VAR-type process are generated from a degree-corrected stochastic co-blockmodel whose community structure evolves over time.
    This is the embedding step that allows separation of sending and receiving roles and low-dimensional summarization.
invented entities (1)
  • latent community paths no independent evidence
    purpose: Time-evolving directed communities that summarize directional dependence in VAR models.
    The target object recovered by the dynamic co-clustering procedure; no independent evidence outside the fitted model is mentioned.

pith-pipeline@v0.9.0 · 5474 in / 1529 out tokens · 86618 ms · 2026-05-10T14:57:32.905271+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 29 canonical work pages

  1. [1]

    Abbe, E. (2018). Community detection and stochastic block models: recent developments. Journal of Machine Learning Research , 18(177):1--86

    2018

  2. [2]

    G., Bollerslev, T., Diebold, F

    Andersen, T. G., Bollerslev, T., Diebold, F. X., and Labys, P. (2003). Modeling and forecasting realized volatility. Econometrica , 71(2):579--625. DOI: https://doi.org/10.1111/1468-0262.00418

    work page doi:10.1111/1468-0262.00418 2003

  3. [3]

    A., and Pipiras, V

    Baek, C., Davis, R. A., and Pipiras, V. (2018). Periodic dynamic factor models: Estimation approaches and applications. Electronic Journal of Statistics , 12(2):4377--4411. DOI: 10.1214/18-EJS1518

    work page doi:10.1214/18-ejs1518 2018

  4. [4]

    and Park, M

    Baek, C. and Park, M. (2021). Sparse vector heterogeneous autoregressive modeling for realized volatility. Journal of the Korean Statistical Society , 50(2):495--510. DOI: 10.1007/s42952-020-00090-5

    work page doi:10.1007/s42952-020-00090-5 2021

  5. [5]

    and Ng, S

    Bai, J. and Ng, S. (2007). Determining the number of primitive shocks in factor models. Journal of Business & Economic Statistics , 25(1):52--60. DOI: 10.1198/073500106000000413

    work page doi:10.1198/073500106000000413 2007

  6. [6]

    Barigozzi, M., Cho, H., and Owens, D. (2024). FNETS: factor-adjusted network estimation and forecasting for high-dimensional time series. Journal of Business & Economic Statistics , 42(3):890--902. DOI: 10.1080/07350015.2023.2257270

    work page doi:10.1080/07350015.2023.2257270 2024

  7. [7]

    Basu, S., Das, S., Michailidis, G., and Purnanandam, A. (2024). A high-dimensional approach to measure connectivity in the financial sector. The Annals of Applied Statistics , 18(2):922--945. DOI: 10.1214/22-AOAS1702

    work page doi:10.1214/22-aoas1702 2024

  8. [8]

    and Michailidis, G

    Basu, S. and Michailidis, G. (2015). Regularized estimation in sparse high-dimensional time series models. The Annals of Statistics , 43(4):1535--1567. DOI: 10.1214/15-AOS1315

    work page doi:10.1214/15-aos1315 2015

  9. [9]

    A fast iterative shrinkage-thresholding algorithm for linear inverse problems

    Beck, A. and Teboulle, M. (2009). A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM Journal on Imaging Sciences , 2(1):183--202. DOI: 10.1137/080716542

    work page doi:10.1137/080716542 2009

  10. [10]

    and Ursu, E

    Boubacar Ma \" nassara, Y. and Ursu, E. (2023). Estimating weak periodic vector autoregressive time series. Test , 32(3):958--997. DOI: 10.1007/s11749-023-00859-w

    work page doi:10.1007/s11749-023-00859-w 2023

  11. [11]

    , year =

    Brownlees, C., Gu mundsson, G. S., and Lugosi, G. (2022). Community detection in partial correlation network models. Journal of Business & Economic Statistics , 40(1):216--226. DOI: 10.1080/07350015.2020.1798241

    work page doi:10.1080/07350015.2020.1798241 2022

  12. [12]

    Brownlees, C., Nualart, E., and Sun, Y. (2018). Realized networks. Journal of Applied Econometrics , 33(7):986--1006. DOI: 10.1002/jae.2642

    work page doi:10.1002/jae.2642 2018

  13. [13]

    Chaudhuri, K., Chung, F., and Tsiatas, A. (2012). Spectral clustering of graphs with general degrees in the extended planted partition model. In Proceedings of the 25th Annual Conference on Learning Theory , volume 23 of JMLR Workshop and Conference Proceedings , pages 35.1--35.23

    2012

  14. [14]

    Corsi, F. (2009). A simple approximate long-memory model of realized volatility. Journal of Financial Econometrics , 7(2):174--196. DOI: 10.1093/jjfinec/nbp001

    work page doi:10.1093/jjfinec/nbp001 2009

  15. [15]

    and Kahan, W

    Davis, C. and Kahan, W. M. (1970). The rotation of eigenvectors by a perturbation. iii. SIAM Journal on Numerical Analysis , 7(1):1--46. DOI: 10.1137/0707001

    work page doi:10.1137/0707001 1970

  16. [16]

    Granger, C. W. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica , 37(3):424--438. DOI: 10.2307/1912791

    work page doi:10.2307/1912791 1969

  17. [17]

    Gu mundsson, G. S. (2025). Detecting giver and receiver spillover groups in large vector autoregressions. Journal of Business & Economic Statistics , pages 1--12. DOI: 10.1080/07350015.2025.2526430

    work page doi:10.1080/07350015.2025.2526430 2025

  18. [18]

    Gu mundsson, G. S. and Brownlees, C. (2021). Detecting groups in large vector autoregressions. Journal of Econometrics , 225(1):2--26. DOI: 10.1016/j.jeconom.2021.03.012

    work page doi:10.1016/j.jeconom.2021.03.012 2021

  19. [19]

    and Baek, C

    Lee, S. and Baek, C. (2023). Volatility changes in cryptocurrencies: evidence from sparse VHAR - MGARCH model. Applied Economics Letters , 30(11):1496--1504. DOI: 10.1080/13504851.2022.2064417

    work page doi:10.1080/13504851.2022.2064417 2023

  20. [20]

    Liu, F., Choi, D., Xie, L., and Roeder, K. (2018). Global spectral clustering in dynamic networks. Proceedings of the National Academy of Sciences , 115(5):927--932. DOI: 10.1073/pnas.1718449115

    work page doi:10.1073/pnas.1718449115 2018

  21. [21]

    L \"u tkepohl, H. (2005). New Introduction to Multiple Time Series Analysis . Springer Science & Business Media. DOI: 10.1007/978-3-540-27752-1

    work page doi:10.1007/978-3-540-27752-1 2005

  22. [22]

    GICS: global industry classification standard

    MSCI (2025). GICS: global industry classification standard. Accessed: 8 January 2025

    2025

  23. [23]

    u ller, U. A., Dacorogna, M. M., Dav \'e , R. D., Olsen, R. B., Pictet, O. V., and Von Weizs \

    M \"u ller, U. A., Dacorogna, M. M., Dav \'e , R. D., Olsen, R. B., Pictet, O. V., and Von Weizs \"a cker, J. E. (1997). Volatilities of different time resolutions—analyzing the dynamics of market components. Journal of Empirical Finance , 4(2-3):213--239. DOI: 10.1016/S0927-5398(97)00007-8

    work page doi:10.1016/s0927-5398(97)00007-8 1997

  24. [24]

    and Rohe, K

    Qin, T. and Rohe, K. (2013). Regularized spectral clustering under the degree-corrected stochastic blockmodel. In Advances in Neural Information Processing Systems , volume 26, pages 3120--3128

    2013

  25. [25]

    Rohe, K., Chatterjee, S., and Yu, B. (2011). Spectral clustering and the high-dimensional stochastic blockmodel. The Annals of Statistics , 39(4):1878--1915. DOI: 10.1214/11-AOS887

    work page doi:10.1214/11-aos887 2011

  26. [26]

    Rohe, K., Qin, T., and Yu, B. (2016). Co-clustering directed graphs to discover asymmetries and directional communities. Proceedings of the National Academy of Sciences , 113(45):12679--12684. DOI: 10.1073/pnas.1525793113

    work page doi:10.1073/pnas.1525793113 2016

  27. [27]

    Sch \"o nemann, P. H. (1966). A generalized solution of the orthogonal Procrustes problem. Psychometrika , 31(1):1--10. DOI: 10.1007/BF02289451

    work page doi:10.1007/bf02289451 1966

  28. [28]

    and Duchesne, P

    Ursu, E. and Duchesne, P. (2009). On modelling and diagnostic checking of vector periodic autoregressive time series models. Journal of Time Series Analysis , 30(1):70--96. DOI: 10.1111/j.1467-9892.2008.00601.x

    work page doi:10.1111/j.1467-9892.2008.00601.x 2009

  29. [29]

    Vershynin, R. (2018). High-Dimensional Probability: An Introduction with Applications in Data Science , volume 47. Cambridge University Press. DOI: 10.1017/9781108231596

    work page doi:10.1017/9781108231596 2018

  30. [30]

    X., Epps, J., and Bailey, J

    Vinh, N. X., Epps, J., and Bailey, J. (2009). Information theoretic measures for clusterings comparison: Is a correction for chance necessary? In Proceedings of the 26th Annual International Conference on Machine Learning , pages 1073--1080. DOI: 10.1145/1553374.1553511

    work page doi:10.1145/1553374.1553511 2009

  31. [31]

    Wang, Z., Liang, Y., and Ji, P. (2020). Spectral algorithms for community detection in directed networks. Journal of Machine Learning Research , 21(153):1--45

    2020

  32. [32]

    C., Li, Z., and Tewari, A

    Wong, K. C., Li, Z., and Tewari, A. (2020). Lasso guarantees for -mixing heavy-tailed time series. The Annals of Statistics , 48(2):1124--1142. DOI: 10.1214/19-AOS1840

    work page doi:10.1214/19-aos1840 2020

  33. [33]

    C., Comin, C

    Ye, C., Wilson, R. C., Comin, C. H., Costa, L. d. F., and Hancock, E. R. (2014). Approximate von Neumann entropy for directed graphs. Physical Review E , 89(5):052804. DOI: 10.1103/PhysRevE.89.052804

    work page doi:10.1103/physreve.89.052804 2014

  34. [34]

    Yin, H., Safikhani, A., and Michailidis, G. (2023). A general modeling framework for network autoregressive processes. Technometrics , pages 1--11. DOI: 10.1080/00401706.2023.2203184

    work page doi:10.1080/00401706.2023.2203184 2023