pith. sign in

arxiv: 2605.22025 · v1 · pith:KDC7WJODnew · submitted 2026-05-21 · 📊 stat.ME

Testing for Serial Independence via Auto Hilbert-Schmidt Independence Criterion

Muyi Li , Yuqing Xu , Zhou Zhou This is my paper

Pith reviewed 2026-05-22 04:21 UTC · model grok-4.3

classification 📊 stat.ME
keywords serial independenceHilbert-Schmidt independence criterionAutoHSICwild bootstrapU-statisticnonlinear dependencestationary time series
0
0 comments X

The pith

AutoHSIC tests serial independence in stationary time series by measuring kernel dependence between an observation and its lag, with a wild bootstrap supplying critical values.

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

The paper develops a Hilbert-Schmidt independence criterion adapted to time series, called AutoHSIC, that checks whether each observation is independent of its lagged counterpart. This produces a U-statistic from overlapping pairs, so the statistic remains dependent even when the null of serial independence holds, unlike standard i.i.d. settings. Asymptotics are derived for both single-lag and portmanteau versions under the null and under fixed alternatives. Because the null limit is non-pivotal, a wild bootstrap is constructed and proved to deliver asymptotically valid critical values. The same framework is extended to residual diagnostics after model fitting and is illustrated on multivariate, functional, and matrix-valued series.

Core claim

The AutoHSIC statistic, formed as a lagged U-statistic on overlapping observations, converges to a non-pivotal limit under the null of serial independence and to a positive constant under fixed alternatives; the wild bootstrap consistently approximates the null distribution, yielding asymptotically valid single-lag and portmanteau tests that extend directly to residual-based checks after parameter estimation.

What carries the argument

AutoHSIC, the lagged Hilbert-Schmidt independence criterion realized as a degenerate U-statistic on temporally overlapping pairs.

If this is right

  • Single-lag and portmanteau tests possess explicit limiting distributions under the null and fixed alternatives.
  • The wild bootstrap delivers critical values whose size and power properties match the asymptotic theory.
  • The procedure remains valid for residual series after fitting a parametric model, with the estimation effect incorporated into the bootstrap.
  • The same construction applies without change to multivariate, functional, and matrix time series.

Where Pith is reading between the lines

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

  • The method could serve as a benchmark when comparing power against linear portmanteau tests or mutual-information procedures on nonlinear processes.
  • Replacing the kernel with a data-driven choice might further improve finite-sample sensitivity to specific forms of dependence.
  • The overlapping-U-statistic structure suggests similar bootstrap corrections could be used for other kernel dependence measures on dependent data.

Load-bearing premise

The time series must be strictly stationary so that the joint distribution of each observation and its lag is time-invariant and the U-statistic has a well-defined limiting law.

What would settle it

If the wild-bootstrap p-values for the AutoHSIC statistic fail to converge to uniform under the i.i.d. null as sample size increases, the claimed asymptotic validity would be refuted.

read the original abstract

We develop a Hilbert--Schmidt independence criterion (HSIC)-based framework for testing serial independence in strictly stationary time series. The proposed auto Hilbert--Schmidt independence criterion (AutoHSIC) measures dependence between an observation and its lagged counterpart, providing a kernel-based approach to detecting nonlinear serial dependence. The empirical AutoHSIC statistic is a lagged U-statistic constructed from overlapping observations, and hence inherits temporal dependence even under the i.i.d. null. Its asymptotic analysis therefore differs from standard i.i.d. HSIC theory and must account for degeneracy under the null. We establish the limiting behaviour of the resulting single-lag and portmanteau tests under the null and under fixed alternatives. Since the limiting null distribution is non-pivotal, we develop a wild bootstrap procedure for critical value approximation and prove its asymptotic validity. The framework is further extended to residual-based model diagnostics, where parameter estimation affects the null distribution. Simulations and empirical applications illustrate its ability to detect nonlinear serial dependence in multivariate, functional and matrix time series.

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

0 major / 3 minor

Summary. The manuscript develops an AutoHSIC statistic based on the Hilbert-Schmidt Independence Criterion to test for serial independence in strictly stationary time series. The empirical version is a lagged U-statistic that accounts for overlapping observations, with derived limiting distributions under the null (degenerate case) and fixed alternatives. A wild bootstrap is introduced to approximate the non-pivotal null distribution and its asymptotic validity is established. The method is extended to residual-based diagnostics after parameter estimation and is illustrated via simulations and applications to multivariate, functional, and matrix time series.

Significance. If the asymptotic derivations and bootstrap consistency hold, the work supplies a flexible nonparametric tool for detecting nonlinear serial dependence that complements linear portmanteau tests. The explicit handling of degeneracy induced by overlapping blocks and the extension to residual diagnostics after estimation are clear strengths, as is the coverage of non-standard data types. These features could make the procedure useful for model checking in a range of time-series settings.

minor comments (3)
  1. The statement of the moment and mixing conditions on the kernel (presumably in the assumptions preceding the main theorems) could be made more explicit with concrete rates to facilitate verification of the bootstrap consistency argument.
  2. In the simulation section, the reported power curves would benefit from the inclusion of Monte Carlo standard errors or confidence bands so that differences across methods can be assessed for statistical significance.
  3. Notation for the estimated residuals in the diagnostic extension should be introduced earlier and kept consistent with the notation used for the raw series to avoid confusion when reading the asymptotic results for the residual-based statistic.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and constructive assessment of our manuscript. We appreciate the recommendation for minor revision and the recognition of the strengths of the AutoHSIC framework, including its handling of degeneracy, extension to residual diagnostics, and applicability to non-standard data types.

Circularity Check

0 steps flagged

No significant circularity; limiting distributions and bootstrap validity derived independently

full rationale

The paper derives the limiting null and alternative distributions for the degenerate lagged U-statistic (AutoHSIC) under strict stationarity, explicitly accounting for overlap-induced dependence and degeneracy. The wild bootstrap is introduced and its asymptotic validity is proven as a separate approximation device for the non-pivotal limit, using residual adjustments for estimated parameters where needed. These steps rely on standard U-statistic theory, mixing conditions, and bootstrap consistency arguments that do not reduce to fitted inputs, self-definitions, or load-bearing self-citations by construction. The residual-based extension similarly adjusts the null distribution via explicit derivation rather than re-using the original statistic's fitted values. The central claims remain self-contained against external benchmarks such as standard HSIC theory and wild bootstrap results for dependent data.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the domain assumption of strict stationarity and on the technical claim that the wild bootstrap consistently approximates the non-pivotal null limit induced by overlapping blocks.

axioms (1)
  • domain assumption The underlying time series is strictly stationary.
    Explicitly required in the abstract for the limiting theory to hold.

pith-pipeline@v0.9.0 · 5704 in / 1130 out tokens · 44425 ms · 2026-05-22T04:21:12.886484+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages

  1. [1]

    The law of large numbers for

    Arcones, Miguel , year=. The law of large numbers for. Electronic Communications in Probability , volume=

    work page

  2. [2]

    Journal of Econometrics , volume=

    Generalized autoregressive conditional heteroskedasticity , author=. Journal of Econometrics , volume=. 1986 , publisher=

    work page 1986

  3. [3]

    Journal of the American Statistical Association , volume=

    Distribution of residual autocorrelations in autoregressive-integrated moving average time series models , author=. Journal of the American Statistical Association , volume=. 1970 , publisher=

    work page 1970

  4. [4]

    Mixing and moment properties of various

    Marine Carrasco and Xiaohong Chen , journal =. Mixing and moment properties of various

    work page

  5. [5]

    Functional

    Cerovecki, Cl. Functional. Journal of Econometrics , volume=. 2019 , publisher=

    work page 2019

  6. [6]

    An approach to

    Cheng, Bing and Tong, Howell , journal=. An approach to. 2026 , publisher=

    work page 2026

  7. [7]

    International Conference on Machine Learning , pages=

    A kernel independence test for random processes , author=. International Conference on Machine Learning , pages=. 2014 , organization=

    work page 2014

  8. [8]

    Journal of Financial Econometrics , volume =

    Corsi, Fulvio , title =. Journal of Financial Econometrics , volume =

    work page

  9. [9]

    Bernoulli , volume=

    Applications of distance correlation to time series , author=. Bernoulli , volume=. 2018 , publisher=

    work page 2018

  10. [10]

    1963 , publisher=

    Linear Operators Part 2: Spectral Theory , author=. 1963 , publisher=

    work page 1963

  11. [11]

    Engle , title =

    Robert F. Engle , title =. Econometrica , volume =. 1982 , publisher=

    work page 1982

  12. [12]

    Multivariate simultaneous generalized

    Engle, Robert F and Kroner, Kenneth F , journal=. Multivariate simultaneous generalized. 1995 , publisher=

    work page 1995

  13. [13]

    Journal of the American Statistical Association , volume=

    Goodness-of-fit tests for linear and nonlinear time series models , author=. Journal of the American Statistical Association , volume=. 2006 , publisher=

    work page 2006

  14. [14]

    2003 , publisher=

    Nonlinear Time Series: Nonparametric and Parametric Methods , author=. 2003 , publisher=

    work page 2003

  15. [15]

    International Statistical Review/Revue Internationale de Statistique , pages=

    A consistent test for bivariate dependence , author=. International Statistical Review/Revue Internationale de Statistique , pages=. 1993 , publisher=

    work page 1993

  16. [16]

    Technometrics , volume=

    Consistent testing for pairwise dependence in time series , author=. Technometrics , volume=. 2017 , publisher=

    work page 2017

  17. [17]

    Kernel measures of conditional dependence , year =

    Fukumizu, Kenji and Gretton, Arthur and Sun, Xiaohai and Sch\". Kernel measures of conditional dependence , year =. Proceedings of the 21st International Conference on Neural Information Processing Systems , pages =

    work page

  18. [18]

    Characteristic kernels on groups and semigroups , year =

    Fukumizu, Kenji and Sriperumbudur, Bharath and Gretton, Arthur and Sch\'. Characteristic kernels on groups and semigroups , year =. Proceedings of the 22nd International Conference on Neural Information Processing Systems , pages =

    work page

  19. [19]

    Journal of the American Statistical Association , volume=

    Tests for error correlation in the functional linear model , author=. Journal of the American Statistical Association , volume=. 2010 , publisher=

    work page 2010

  20. [20]

    Resampling-free inference for time series via

    Ghoshal, Deep and Shao, Xiaofeng , journal=. Resampling-free inference for time series via. 2026 , publisher=

    work page 2026

  21. [21]

    Biometrika , volume=

    Entropy testing for nonlinear serial dependence in time series , author=. Biometrika , volume=. 2015 , publisher=

    work page 2015

  22. [22]

    Measuring statistical dependence with

    Gretton, Arthur and Bousquet, Olivier and Smola, Alex and Sch. Measuring statistical dependence with. International Conference on Algorithmic Learning Theory , pages=. 2005 , organization=

    work page 2005

  23. [23]

    A kernel statistical test of independence , year =

    Gretton, Arthur and Fukumizu, Kenji and Teo, Choon Hui and Song, Le and Sch\". A kernel statistical test of independence , year =. Proceedings of the 21st International Conference on Neural Information Processing Systems , pages =

    work page

  24. [24]

    Journal of Financial Econometrics , volume =

    Hecq, Alain and Margaritella, Luca and Smeekes, Stephan , title =. Journal of Financial Econometrics , volume =

    work page

  25. [25]

    Econometrica , year=

    Consistent Testing for Serial Correlation of Unknown Form , author=. Econometrica , year=

    work page

  26. [26]

    Journal of the American Statistical Association , volume=

    Hypothesis testing in time series via the empirical characteristic function: A generalized spectral density approach , author=. Journal of the American Statistical Association , volume=. 1999 , publisher=

    work page 1999

  27. [27]

    working paper , year=

    Generalized spectral test for functional martingale difference hypothesis with bootstrap approximation , author=. working paper , year=

    work page

  28. [28]

    A functional version of the

    Hormann, Siegfried and Horvath, Lajos and Reeder, Ron , year =. A functional version of the. Econometric Theory , publisher=

    work page

  29. [29]

    Asymptotic normality for m-dependent and constrained

    Janson, Svante , journal=. Asymptotic normality for m-dependent and constrained. 2023 , publisher=

    work page 2023

  30. [30]

    Biometrika , volume=

    Testing serial independence of object-valued time series , author=. Biometrika , volume=. 2024 , publisher=

    work page 2024

  31. [31]

    Computational Statistics & Data Analysis , volume=

    A kernel-based measure for conditional mean dependence , author=. Computational Statistics & Data Analysis , volume=. 2021 , publisher=

    work page 2021

  32. [32]

    1990 , publisher=

    U-statistics: Theory and Practice , author=. 1990 , publisher=

    work page 1990

  33. [33]

    2003 , author =

    Consistent specification tests for semiparametric/nonparametric models based on series estimation methods , journal =. 2003 , author =

    work page 2003

  34. [34]

    Biometrika , volume=

    On a measure of lack of fit in time series models , author=. Biometrika , volume=. 1978 , publisher=

    work page 1978

  35. [35]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Kernel-based tests for joint independence , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2018 , publisher=

    work page 2018

  36. [36]

    1985 , author =

    Some mixing properties of time series models , journal =. 1985 , author =

    work page 1985

  37. [37]

    Politis , journal =

    Dimitris N. Politis , journal =. The Impact of Bootstrap Methods on Time Series Analysis , volume =

    work page

  38. [38]

    Equivalence of distance-based and

    Sejdinovic, Dino and Sriperumbudur, Bharath and Gretton, Arthur and Fukumizu, Kenji , journal =. Equivalence of distance-based and. 2013 , pages =

    work page 2013

  39. [39]

    The Bell System Technical Journal , volume=

    A mathematical theory of communication , author=. The Bell System Technical Journal , volume=. 1948 , publisher=

    work page 1948

  40. [40]

    Journal of Complexity , volume =

    Mercer theorem for. Journal of Complexity , volume =. 2005 , author =

    work page 2005

  41. [41]

    Measuring and testing dependence by correlation of distances , volume =

    G. Measuring and testing dependence by correlation of distances , volume =. The Annals of Statistics , number =

    work page

  42. [42]

    Partial distance correlation with methods for dissimilarities , volume =

    G. Partial distance correlation with methods for dissimilarities , volume =. The Annals of Statistics , number =

    work page

  43. [43]

    Journal of Business & Economic Statistics , volume=

    A multivariate generalized autoregressive conditional heteroscedasticity model with time-varying correlations , author=. Journal of Business & Economic Statistics , volume=. 2002 , publisher=

    work page 2002

  44. [44]

    Journal of Econometrics , volume=

    Goodness-of-fit testing for time series models via distance covariance , author=. Journal of Econometrics , volume=. 2022 , publisher=

    work page 2022

  45. [45]

    Wang, Guochang and Li, Wai Keung and Zhu, Ke , journal=. New. 2021 , publisher=

    work page 2021

  46. [46]

    2001 , publisher=

    Asymptotic Theory for Econometricians , author=. 2001 , publisher=

    work page 2001

  47. [47]

    Journal of Multivariate Analysis , volume =

    Test of conditional independence in factor models via. Journal of Multivariate Analysis , volume =. 2024 , author =

    work page 2024

  48. [48]

    Limiting behavior of

    Yoshihara, Ken-ichi , journal=. Limiting behavior of. 1976 , publisher=

    work page 1976

  49. [49]

    Yu, Cheng and Li, Dong and Jiang, Feiyu and Zhu, Ke , journal=. Matrix. 2025 , publisher=

    work page 2025

  50. [50]

    Kernel-based conditional independence test and application in causal discovery , booktitle =

    Zhang, Kun and Peters, Jonas and Janzing, Dominik and Sch. Kernel-based conditional independence test and application in causal discovery , booktitle =. 2011 , pages =

    work page 2011

  51. [51]

    Proceedings of the 37th International Conference on Neural Information Processing Systems , pages=

    Zhang, Tao and Zhang, Yaowu and Zhou, Tingyou , title =. Proceedings of the 37th International Conference on Neural Information Processing Systems , pages=. 2023 , publisher =

    work page 2023

  52. [52]

    2018 , journal=

    Conditional mean and quantile dependence testing in high dimension , author=. 2018 , journal=

    work page 2018

  53. [53]

    Journal of Time Series Analysis , volume=

    Measuring nonlinear dependence in time-series, a distance correlation approach , author=. Journal of Time Series Analysis , volume=. 2012 , publisher=

    work page 2012

  54. [54]

    Distance-based and

    Zhu, Changbo and Zhang, Xianyang and Yao, Shun and Shao, Xiaofeng , journal=. Distance-based and. 2020 , publisher=

    work page 2020