pith. sign in

arxiv: 2606.00858 · v1 · pith:CPGT3VFGnew · submitted 2026-05-30 · 📊 stat.ME · math.ST· stat.TH

Change-Point Detection for Object-valued Time Series

Yi Zhang , Changbo Zhu , Xiaofeng Shao This is my paper

Pith reviewed 2026-06-28 18:07 UTC · model grok-4.3

classification 📊 stat.ME math.STstat.TH
keywords change-point detectionobject-valued time seriesself-normalizationmetric spacewild binary segmentationpairwise distancesnonparametric inferencedistributional data
0
0 comments X

The pith

A self-normalization statistic detects shifts in the marginal distribution of object-valued time series and supports consistent multiple change-point estimation.

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

The paper develops a self-normalization based test for change points in the distribution of time series that take values in general metric spaces. The statistic applies to diverse object data such as distributions and networks, handles weak dependence, uses only pairwise distances, and requires almost no tuning parameters. It possesses a pivotal limiting distribution under the null hypothesis. When paired with the wild binary segmentation algorithm, the approach delivers the first consistency results for estimating multiple change points in a nonparametric framework for these series.

Core claim

The central discovery is that the proposed self-normalization statistic for detecting a single change in the marginal distribution has a pivotal limiting null distribution and is consistent under local alternatives. Furthermore, combining this statistic with the wild binary segmentation procedure yields consistent estimation of the number and locations of multiple change points for a broad nonparametric class of weakly dependent object-valued time series.

What carries the argument

The self-normalization (SN) statistic based on cumulative sums of pairwise distances between observations, which normalizes the test statistic to achieve a pivotal limit without tuning parameters.

If this is right

  • The test applies directly to distributional data and network-valued data.
  • The method accommodates weak serial dependence without additional adjustments.
  • Multiple change-point locations and count are consistently estimated in a nonparametric setting.
  • The statistic depends only on pairwise distances and is nearly free of tuning parameters.

Where Pith is reading between the lines

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

  • The framework may extend to change detection in other complex data types like images or shapes stored in metric spaces.
  • Real-time monitoring applications could benefit from the tuning-free nature for streaming object data.
  • Similar self-normalization ideas might improve robustness in existing change-point methods for Euclidean data.

Load-bearing premise

The observations form a weakly dependent time series taking values in a metric space.

What would settle it

A simulation study with known change points in a metric-space valued series where the self-normalization statistic does not converge to its claimed pivotal null distribution or where wild binary segmentation fails to recover the true number and locations consistently.

Figures

Figures reproduced from arXiv: 2606.00858 by Changbo Zhu, Xiaofeng Shao, Yi Zhang.

Figure 1
Figure 1. Figure 1: Size adjusted power when testing for change point in marginal distribution of [PITH_FULL_IMAGE:figures/full_fig_p020_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Size adjusted power for testing hypothesis on the existence of a change point for [PITH_FULL_IMAGE:figures/full_fig_p023_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Size adjusted power for testing hypothesis on the existence of a change point for [PITH_FULL_IMAGE:figures/full_fig_p024_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Heat plot of 96 density functions. Each column represents a density function. [PITH_FULL_IMAGE:figures/full_fig_p027_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of weighted undirected taxi-trip networks indexed by date. [PITH_FULL_IMAGE:figures/full_fig_p028_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: For each node i ∈ {Manhattan, Queens, EWR,Brooklyn,Bronx, Staten Island}, the edge weights {wij,t : j ̸= i} are plotted as multivariate time series. 29 [PITH_FULL_IMAGE:figures/full_fig_p029_6.png] view at source ↗
read the original abstract

This article is concerned with change point detection for object-valued data that reside in a metric space, which has attracted some recent interests in statistics and econometrics literature. The existing methods either focus on independent data or can only detect change in the Fr\'echet mean or variance. In this paper, we propose a self-normalization (SN, hereafter) based statistic for detecting a shift in the marginal distribution of object-valued time series. Our test is universally applicable to a wide range of object-valued data, such as distributional and network data, and can accommodate weak serial dependence. In addition the proposed test statistic is almost tuning parameter free, has pivotal limiting null distribution and only uses the pairwise distances. When combined with the Wild Binary Segmentation algorithm (WBS, hereafter), our statistic can be used to estimate the number and locations of multiple change points. Asymptotic results for our SN based statistic are derived under both null and local alternatives in the single change point setting. For the first time, the WBS estimation consistency is shown for a broad class of object-valued time series and in a nonparametric setting, which requires new non-standard theoretical arguments. Extensive numerical experiments and real data analysis are conducted to illustrate the effectiveness and broad applicability of our proposed method.

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 paper proposes a self-normalization (SN) based test statistic for detecting shifts in the marginal distribution of weakly dependent object-valued time series taking values in a general metric space. The statistic relies solely on pairwise distances, is nearly tuning-parameter free, and is shown to have a pivotal limiting null distribution. When combined with the wild binary segmentation (WBS) algorithm, it yields consistent estimation of the number and locations of multiple change points. Asymptotic results are derived for the single-change case under the null and local alternatives, and new arguments establish WBS consistency in this nonparametric setting. Numerical experiments and real-data examples illustrate performance on distributional and network data.

Significance. If the asymptotic and consistency results hold, the work supplies a general, distance-based framework for change-point detection that extends beyond Fréchet mean or variance changes and applies to dependent non-Euclidean data. The pairwise-distance construction and the first nonparametric consistency proof for WBS in this class are clear strengths; the paper also supplies reproducible numerical evidence and real-data illustrations.

minor comments (3)
  1. [§2.2] §2.2: the definition of the SN statistic could explicitly state the normalization constant before the limiting-distribution theorem to improve readability.
  2. [Theorem 3.1] Theorem 3.1: the local-alternative rate is stated but the precise form of the drift term in the limiting process is not displayed; adding the explicit expression would clarify the power analysis.
  3. [Figure 4] Figure 4: axis labels and legend entries are too small for print; increasing font size would aid interpretation of the WBS segmentation results.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive review, accurate summary of the contributions, and recommendation of minor revision. We are pleased that the general distance-based framework, the tuning-parameter-free property, and the first nonparametric WBS consistency results are recognized as strengths.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation constructs the SN statistic directly from pairwise distances between observations in a general metric space and derives its pivotal null limit and local-alternative behavior under weak dependence. The WBS consistency result is stated to require new nonparametric arguments rather than reducing to a fitted parameter, a self-citation chain, or an ansatz imported from prior work by the same authors. No load-bearing step equates a claimed prediction or uniqueness result to its own inputs by construction; the central claims remain independent of the fitted values or self-referential definitions.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on standard domain assumptions for metric spaces and weak dependence; no free parameters or invented entities are mentioned in the abstract.

axioms (2)
  • domain assumption Observations reside in a metric space equipped with a distance function.
    Explicitly stated as the setting for object-valued data.
  • domain assumption The time series satisfies weak serial dependence.
    Required for the method to accommodate dependence and for the asymptotic theory.

pith-pipeline@v0.9.1-grok · 5749 in / 1228 out tokens · 24021 ms · 2026-06-28T18:07:24.893996+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

111 extracted references · 5 canonical work pages

  1. [1]

    Biometrika , volume=

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

  2. [2]

    Journal of Econometrics , volume=

    Spherical autoregressive models, with application to distributional and compositional time series , author=. Journal of Econometrics , volume=

  3. [3]

    Journal of the American Statistical Association , volume=

    Self-normalization for time series: a review of recent developments , author=. Journal of the American Statistical Association , volume=

  4. [4]

    Journal of Statistical Computation and Simulation , volume=

    Spectral density bandwidth choice: source of nonmonotonic power for tests of a mean shift in a time series , author=. Journal of Statistical Computation and Simulation , volume=

  5. [5]

    Journal of Machine Learning Research , volume=

    Network regression with graph Laplacians , author=. Journal of Machine Learning Research , volume=

  6. [6]

    2022 , note =

    BatchGetSymbols: Downloads and Organizes Financial Data for Multiple Tickers , author =. 2022 , note =

    2022

  7. [7]

    arXiv preprint arXiv:2105.08976 , year=

    High-dimensional change-point detection using generalized homogeneity metrics , author=. arXiv preprint arXiv:2105.08976 , year=

    arXiv

  8. [8]

    Journal of the American Statistical Association , volume=

    Testing for Change Points in Time Series , author=. Journal of the American Statistical Association , volume=. 2010 , publisher=

    2010

  9. [9]

    Malliavin Calculus and Stochastic Analysis: A Festschrift in Honor of David Nualart , pages=

    Large deviations for Hilbert-space-valued Wiener processes: A sequence space approach , author=. Malliavin Calculus and Stochastic Analysis: A Festschrift in Honor of David Nualart , pages=. 2013 , organization=

    2013

  10. [10]

    Change-point detection and bootstrap for Hilbert space valued random fields , journal =

    Béatrice Bucchia and Martin Wendler , keywords =. Change-point detection and bootstrap for Hilbert space valued random fields , journal =. 2017 , issn =. doi:https://doi.org/10.1016/j.jmva.2017.01.007 , url =

    work page doi:10.1016/j.jmva.2017.01.007 2017

  11. [11]

    Bradley , title =

    Richard C. Bradley , title =. Probability Surveys , publisher =. 2005 , doi =

    2005

  12. [12]

    EQUIVALENCE OF DISTANCE-BASED AND RKHS-BASED STATISTICS IN HYPOTHESIS TESTING , urldate =

    Dino Sejdinovic and Bharath Sriperumbudur and Arthur Gretton and Kenji Fukumizu , journal =. EQUIVALENCE OF DISTANCE-BASED AND RKHS-BASED STATISTICS IN HYPOTHESIS TESTING , urldate =

  13. [13]

    Sequential block bootstrap in a Hilbert space with application to change point analysis , urldate =

    Sharipov, Olimjon and Tewes, Johannes and Wendler, Martin , journal =. Sequential block bootstrap in a Hilbert space with application to change point analysis , urldate =

  14. [14]

    Matteson and Nicholas A

    David S. Matteson and Nicholas A. James , journal =. A Nonparametric Approach for Multiple Change Point Analysis of Multivariate Data , urldate =

  15. [15]

    Electronic Journal of Statistics , volume=

    A new framework for distance and kernel-based metrics in high dimensions , author=. Electronic Journal of Statistics , volume=. 2021 , publisher=

    2021

  16. [16]

    The Annals of Probability , number =

    Russell Lyons , title =. The Annals of Probability , number =. 2021 , doi =

    2021

  17. [17]

    LECTURE NOTES IN MATHEMATICS-SPRINGER-VERLAG- , volume=

    Separability and completeness for the Wasserstein distance , author=. LECTURE NOTES IN MATHEMATICS-SPRINGER-VERLAG- , volume=. 2008 , publisher=

    2008

  18. [18]

    The Annals of Probability , number =

    Russell Lyons , title =. The Annals of Probability , number =. 2013 , doi =

    2013

  19. [19]

    WILD BINARY SEGMENTATION FOR MULTIPLE CHANGE-POINT DETECTION , urldate =

    Piotr Fryzlewicz , journal =. WILD BINARY SEGMENTATION FOR MULTIPLE CHANGE-POINT DETECTION , urldate =

  20. [20]

    The Annals of Probability , pages=

    Rates of convergence for empirical processes of stationary mixing sequences , author=. The Annals of Probability , pages=. 1994 , publisher=

    1994

  21. [21]

    Dubey, Paromita and M. Fr. The Annals of Statistics , volume=. 2020 , publisher=

    2020

  22. [22]

    Weak convergence and empirical processes , pages=

    Weak convergence , author=. Weak convergence and empirical processes , pages=. 1996 , publisher=

    1996

  23. [23]

    Scandinavian Journal of Statistics , volume=

    Weak Convergence of the Regularization Path in Penalized M-Estimation , author=. Scandinavian Journal of Statistics , volume=. 2010 , publisher=

    2010

  24. [24]

    Zeitschrift f

    Invariance principles for sums of Banach space valued random elements and empirical processes , author=. Zeitschrift f. 1983 , publisher=

    1983

  25. [25]

    Colloquia Mathematica Societatis J

    Invariance principles for sums of mixing random elements and the multivariate empirical process , author=. Colloquia Mathematica Societatis J

  26. [26]

    Econometric Theory , volume=

    New ways to prove central limit theorems , author=. Econometric Theory , volume=. 1985 , publisher=

    1985

  27. [27]

    Journal of mathematical analysis and applications , volume=

    A comparison inequality for sums of independent random variables , author=. Journal of mathematical analysis and applications , volume=. 2001 , publisher=

    2001

  28. [28]

    The Annals of Statistics , volume=

    Weakly dependent functional data , author=. The Annals of Statistics , volume=. 2010 , publisher=

    2010

  29. [29]

    Stochastic Processes and their Applications , volume=

    Weak invariance principles for sums of dependent random functions , author=. Stochastic Processes and their Applications , volume=. 2013 , publisher=

    2013

  30. [30]

    arXiv preprint arXiv:1107.3806 , year=

    Asymptotics for minimisers of convex processes , author=. arXiv preprint arXiv:1107.3806 , year=

    Pith/arXiv arXiv

  31. [31]

    Dubey, Paromita and M. Fr. Biometrika , volume=. 2019 , publisher=

    2019

  32. [32]

    Dryden, Ian L and Koloydenko, Alexey and Zhou, Diwei , journal=. Non-. 2009 , publisher=

    2009

  33. [33]

    The Annals of Applied Statistics , volume=

    Hypothesis testing for network data in functional neuroimaging , author=. The Annals of Applied Statistics , volume=

  34. [34]

    Hyndman , title =

    Han Lin Shang and Rob J. Hyndman , title =. Journal of Computational and Graphical Statistics , volume =. 2017 , publisher =

    2017

  35. [35]

    Journal of the Royal Statistical Society: Series A (Statistics in Society) , volume =

    Mazzuco, Stefano and Scarpa, Bruno , title =. Journal of the Royal Statistical Society: Series A (Statistics in Society) , volume =

  36. [36]

    Petersen, Alexander and M. Fr. The Annals of Statistics , volume=. 2019 , publisher=

    2019

  37. [37]

    and Wu, Yichao and M

    Tucker, Danielle C. and Wu, Yichao and M. Variable Selection for Global Fréchet Regression , journal =. 2023 , publisher =

    2023

  38. [38]

    Dimension reduction for Fr

    Zhang, Qi and Xue, Lingzhou and Li, Bing , journal=. Dimension reduction for Fr. 2023 , publisher=

    2023

  39. [39]

    arXiv preprint arXiv:2105.05439 , year=

    Autoregressive optimal transport models , author=. arXiv preprint arXiv:2105.05439 , year=

    arXiv

  40. [40]

    arXiv preprint arXiv:2203.12783 , year=

    Spherical Autoregressive Models, With Application to Distributional and Compositional Time Series , author=. arXiv preprint arXiv:2203.12783 , year=

    arXiv

  41. [41]

    Journal of Time Series Analysis , volume=

    Wasserstein autoregressive models for density time series , author=. Journal of Time Series Analysis , volume=

  42. [42]

    A self-normalized approach to confidence interval construction in time series , volume =

    Xiaofeng Shao , journal =. A self-normalized approach to confidence interval construction in time series , volume =

  43. [43]

    Scientific reports , volume=

    Network evolution of regional brain volumes in young children reflects neurocognitive scores and mother’s education , author=. Scientific reports , volume=

  44. [44]

    Human Brain Mapping , volume =

    Zhu, Changbo and Chen, Yaqing and Müller, Hans-Georg and Wang, Jane-Ling and O'Muircheartaigh, Jonathan and Bruchhage, Muriel and Deoni, Sean , title =. Human Brain Mapping , volume =

  45. [45]

    arXiv preprint arXiv:2312.15376 , year=

    Geodesic optimal transport regression , author=. arXiv preprint arXiv:2312.15376 , year=

    arXiv

  46. [46]

    The Annals of Statistics , volume=

    Hypothesis testing for high-dimensional time series via self-normalization , author=. The Annals of Statistics , volume=

  47. [47]

    The Annals of Statistics , volume=

    Inference for change points in high dimensional data via self-normalization , author=. The Annals of Statistics , volume=

  48. [48]

    Electronic Journal of Statistics , volume=

    Testing the structural stability of temporally dependent functional observations and application to climate projections , author=. Electronic Journal of Statistics , volume=

  49. [49]

    Bernoulli , volume=

    Two sample inference for the second-order property of temporally dependent functional data , author=. Bernoulli , volume=. 2015 , publisher=

    2015

  50. [50]

    Fr. Les. Annales de l'institut Henri Poincar

  51. [51]

    arXiv preprint arXiv:1903.01598 , year=

    Change-point detection for multivariate and non-euclidean data with local dependency , author=. arXiv preprint arXiv:1903.01598 , year=

    Pith/arXiv arXiv 1903

  52. [52]

    Journal of the American Statistical Association, to appear , year =

    Paromita Dubey and Hans-Georg Müller , title =. Journal of the American Statistical Association, to appear , year =

  53. [53]

    2018 , volume =

    Ting Zhang and Liliya Lavitas , title =. 2018 , volume =

    2018

  54. [54]

    1968 , publisher=

    Convergence of Probability Measures , author=. 1968 , publisher=

    1968

  55. [55]

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

    Functional models for time-varying random objects , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2020 , publisher=

    2020

  56. [56]

    Ying, Chao and Yu, Zhou , year=. Fr

  57. [57]

    Instituto Technologico Autonomo de Mexico, Technical Notes , year=

    Size corrected power for bootstrap tests , author=. Instituto Technologico Autonomo de Mexico, Technical Notes , year=

  58. [58]

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

    Inference for linear models with dependent errors , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2013 , publisher=

    2013

  59. [59]

    Time series analysis of

    Jiang, Feiyu and Zhao, Zifeng and Shao, Xiaofeng , journal=. Time series analysis of. 2020 , publisher=

    2020

  60. [60]

    Modelling the

    Jiang, Feiyu and Zhao, Zifeng and Shao, Xiaofeng , journal=. Modelling the. 2022 , publisher=

    2022

  61. [61]

    Second-order comparison of

    Panaretos, Victor M and Kraus, David and Maddocks, John H , journal=. Second-order comparison of. 2010 , publisher=

    2010

  62. [62]

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

    Detecting and dating structural breaks in functional data without dimension reduction , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2018 , publisher=

    2018

  63. [63]

    Journal of the American Statistical Association , volume=

    Testing that a dependent process is uncorrelated , author=. Journal of the American Statistical Association , volume=. 2001 , publisher=

    2001

  64. [64]

    Econometrica , volume=

    Tests for parameter instability and structural change with unknown change point , author=. Econometrica , volume=

  65. [65]

    https://arxiv.org/abs/2303.10808 , year=

    Dimension-agnostic change point detection , author=. https://arxiv.org/abs/2303.10808 , year=

    arXiv

  66. [66]

    Journal of Royal Statistical Society, Series B , volume=

    Segmenting time series via self-normalization , author=. Journal of Royal Statistical Society, Series B , volume=

  67. [67]

    Annals of Statistics , volume=

    Wild binary segmentation for multiple change-point detection , author=. Annals of Statistics , volume=

  68. [68]

    Biometrika , volume=

    A seeded binary segmentation: A general methodology for fast and optimal change point detection , author=. Biometrika , volume=

  69. [69]

    Journal of Econometrics , volume=

    Testing stationarity of functional time series , author=. Journal of Econometrics , volume=

  70. [70]

    Annals of Statistics , volume=

    Local Linear Quantile Estimation for Non-stationary Time Series , author=. Annals of Statistics , volume=

  71. [71]

    arXiv preprint arXiv:2311.16025 , year=

    Change Point Detection for Random Objects using Distance Profiles , author=. arXiv preprint arXiv:2311.16025 , year=

    arXiv

  72. [72]

    Journal of Time Series Analysis , volume=

    Least squares estimation of a shift in linear processes , author=. Journal of Time Series Analysis , volume=. 1994 , publisher=

    1994

  73. [73]

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

    Zhang, Yi and Shao, Xiaofeng , title = ". Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =. 2024 , issn =. doi:10.1093/jrsssb/qkad108 , url =

    work page doi:10.1093/jrsssb/qkad108 2024

  74. [74]

    E. S. Page , journal =. A Test for a Change in a Parameter Occurring at an Unknown Point , urldate =

  75. [75]

    Journal of Time Series Analysis , volume=

    Structural breaks in time series , author=. Journal of Time Series Analysis , volume=. 2013 , publisher=

    2013

  76. [76]

    Journal of the American Statistical Association , volume=

    Detection of multiple structural breaks in multivariate time series , author=. Journal of the American Statistical Association , volume=. 2015 , publisher=

    2015

  77. [77]

    Selective review of offline change point detection methods , journal =

    Charles Truong and Laurent Oudre and Nicolas Vayatis , keywords =. Selective review of offline change point detection methods , journal =. 2020 , issn =. doi:https://doi.org/10.1016/j.sigpro.2019.107299 , url =

    work page doi:10.1016/j.sigpro.2019.107299 2020

  78. [78]

    Knowledge and information systems , volume=

    A survey of methods for time series change point detection , author=. Knowledge and information systems , volume=. 2017 , publisher=

    2017

  79. [79]

    Journal of the Royal Statistical Society Series A: Statistics in Society , volume=

    Fitting age-specific fertility rates by a flexible generalized skew normal probability density function , author=. Journal of the Royal Statistical Society Series A: Statistics in Society , volume=. 2015 , publisher=

    2015

  80. [80]

    Statistical methods in medical research , volume=

    Glucodensities: A new representation of glucose profiles using distributional data analysis , author=. Statistical methods in medical research , volume=. 2021 , publisher=

    2021

Showing first 80 references.