Pith. sign in

REVIEW 2 major objections 5 minor 232 references

A single moving-window U-statistic framework tests, localizes, and intervals multiple high-dimensional change points for general parameters, including under heavy tails.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

A moving-window two-sample U-statistic framework unifies high-dimensional multiple change-point testing, optimal localization via U-PRA projection, and confidence intervals for general kernels, including heavy-tailed data.

T0 review reviewed 2026-07-14 challenge →

load-bearing objection Solid, usable unification of high-d multiple change-point testing/estimation/inference via moving-window U-statistics; independence is the real scope limit, not a hidden crack in the math. the 2 major comments →

arxiv 2607.11256 v1 pith:QARTWTTO submitted 2026-07-13 stat.ME

A General U-Statistic Framework for High-Dimensional Multiple Change-Point Analysis

Bin Liu , Yufeng Liu This is my paper
classification stat.ME MSC 62H1562G1062G2062F40
keywords change-point analysishigh-dimensional dataU-statisticsmultiplier bootstrapsparse alternativesheavy tailsprojection refinementconfidence intervals
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper claims that high-dimensional multiple change-point analysis need not be reinvented for each parameter (mean, variance, robust location) or each task (test, estimate, interval). By placing a user-chosen antisymmetric two-sample kernel inside a moving window of width G, one obtains a coordinate-wise U-statistic process whose ℓ∞ maximum detects sparse jumps, whose above-threshold peaks give consistent initial locations, and whose projection onto an estimated active direction (U-PRA) yields minimax-optimal localization rates together with an argmax-of-drifted-Brownian-motion limit that produces valid confidence intervals. Because the moment conditions live on the kernel rather than the raw data, rank or sign kernels keep size and coverage under heavy tails and contamination where classical mean-based CUSUM methods fail. The same construction covers variance changes by taking kernels of squares or signs of squares. Numerical experiments and a bladder-tumor aCGH application illustrate that the unified pipeline recovers number, locations, and intervals more reliably than existing specialized competitors, especially when the data are heavy-tailed.

Core claim

Under kernel moment and bandwidth conditions that allow dimension to grow exponentially with sample size, the ℓ∞ moving-window U-statistic test controls Type I error and attains the minimax sparse detection boundary; the initial estimators consistently recover the number of change points with near-optimal localization; and the U-PRA refined estimators achieve the optimal rate |ẽγm − γm| = OP(1/∥θ(m)∥²) with limiting distributions that yield asymptotically valid confidence intervals for general parameters.

What carries the argument

The moving-window two-sample U-statistic Tj(k) built from an antisymmetric kernel h, together with its Hoeffding decomposition that separates a deterministic triangular signal peaking at each true change point from a controllable stochastic remainder; the ℓ∞ aggregation and multiplier bootstrap for testing; and the U-PRA projection of the process onto an estimated active set for optimal refinement and inference.

Load-bearing premise

The observations are independent across time; the paper’s own residual-ACF check and serial-dependence simulations show that even moderate AR(1) dependence already produces extra false peaks and undercovering intervals.

What would settle it

Generate high-dimensional sequences with known sparse mean or variance jumps under Student-t or contaminated noise, run the full multiscale U-PRA pipeline, and check whether empirical Type I error stays near the nominal level, Hausdorff localization error scales as 1/signal-squared, and confidence-interval coverage approaches 95 percent; any systematic size inflation or coverage collapse under the paper’s stated kernels would falsify the claims.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper develops a unified high-dimensional multiple change-point framework based on moving-window two-sample U-statistics with user-chosen antisymmetric kernels. It covers testing (ℓ∞ aggregation with a multiplier bootstrap), estimation of the number and locations of change points (initial MOSUM-type peaks refined by the U-PRA projection algorithm), and inference (argmax limits of drifted weighted Brownian motions yielding confidence intervals). Theory gives size control, power under sparse alternatives, consistency of the initial estimators, optimal localization rates after refinement, and limiting distributions under kernel moment, non-degeneracy, and bandwidth conditions that allow d to grow exponentially in n. Simulations cover mean and variance changes under Gaussian, Student-t, and contaminated errors, plus multiscale bandwidth aggregation and sensitivity checks; an aCGH application and an R package are provided.

Significance. If the claims hold under the stated conditions, the contribution is substantial: a single kernel-based pipeline that unifies testing, estimation, and CI construction for general parameters (mean, variance, robust contrasts) in high dimensions, with explicit minimax-type rates and bootstrap validity, and with documented robustness under heavy tails via bounded kernels. Strengths include multi-step Gaussian-approximation and Hoeffding arguments for Theorems 3.1–3.5, extensive finite-sample evidence (including active-set misspecification and mild dependence), public code, and a clear comparison to related MOSUM work (Zhou et al., 2025). The main scope limitation is temporal independence, which the authors themselves quantify in Appendix H; that does not invalidate the internal theory but bounds applicability to strongly dependent series.

major comments (2)
  1. [§1 model; §3; Appendix H.1; §5] The model in §1 and all Hoeffding/Gaussian-approximation arguments in §3 assume independent observations across time. Appendix H.1 shows that AR(1) dependence with ρ≥0.4 already inflates false positives and degrades CI coverage, and Figure 7 documents residual lag-1 autocorrelation in the aCGH data. The independence premise is therefore load-bearing for the claimed size, rates, and intervals. The paper should state this limitation clearly in the main text (Introduction or Discussion), not only in the appendix, and qualify the aCGH conclusions accordingly (e.g., as exploratory under mild short-range dependence).
  2. [Abstract; Theorem 3.2; Theorem 3.5] The abstract and §1 claim minimax-optimal power and localization rates. Theorems 3.2 and 3.5 establish matching upper bounds of the usual sparse high-dimensional form, but the manuscript does not state or prove matching lower bounds, nor does it cite a specific lower-bound result that applies to the general U-statistic kernel setting. Please either add a short lower-bound argument/citation tailored to the kernel-based sparse alternative, or rephrase “minimax-optimal” to “rate-optimal under standard sparse high-dimensional benchmarks” where the comparison is only to known rates for mean-based problems.
minor comments (5)
  1. [§2.4] Figure 2 caption refers to “Figure?? (b)” earlier in the text; fix the broken cross-reference.
  2. [Throughout] Typos and wording: “includeing” (p. 3), “Appdenx” (p. 20), “refinemet” (p. 19), “dimen-sions” (Table 4 header), “defalut” (p. 24). A careful proofread would help.
  3. [§3.1.1; Appendix D] Appendix D verifies Assumptions (A.1)–(A.3) for four kernels; a short pointer in the main text (e.g., after stating (A.1)–(A.4)) would help readers who only skim the theory section.
  4. [§2.6; Appendix F] Default tuning (η=0.15, ρ=0.1, multiscale G grid) is described in Appendix F; a one-sentence summary in §2.6 would improve reproducibility for readers who do not open the supplement.
  5. [Table 2] In Table 2 and related estimation tables, LZZL and Inspect are only reported for a single bandwidth column; a brief note that those methods do not use G would avoid confusion.

Circularity Check

0 steps flagged

No significant circularity: rates, power, and limiting distributions are derived from stated kernel moments, non-degeneracy, and bandwidth conditions rather than fitted inputs renamed as predictions.

full rationale

The paper’s central claims (size control of the ℓ∞ moving-window U-statistic test, minimax-optimal power under sparse alternatives, consistency of the initial estimators, optimal localization of U-PRA refinements, and argmax-of-drifted-Brownian-motion limits for confidence intervals) are obtained from Hoeffding decompositions of a user-specified antisymmetric kernel under explicit Assumptions (A.1)–(A.4), (B.1)–(B.3), and (C.1)–(C.2). Bootstrap critical values and the data-driven active-set threshold w+ are calibrated from the observed sample, but they are not used to define the population parameters θ(m) or the localization rates being estimated; they only implement the procedures whose asymptotic properties are proved under those population conditions. Self-citations (e.g., to Liu et al. 2020, Eichinger & Kirch 2018, Chernozhukov et al. 2017) supply technical tools or special cases and do not force the present theorems by construction. The independence assumption is a scope limitation, not a circular step. Score 1 reflects only routine methodological self-citation that is not load-bearing for the claimed derivations.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 1 invented entities

The central claims rest on classical independence and moment conditions for U-statistics plus several tuning constants that are either fixed by default or chosen data-driven; no new physical entities are postulated. The free parameters and domain assumptions listed below are exactly those needed for the size, power, consistency, and limiting-distribution theorems.

free parameters (5)
  • bandwidth G (or multiscale grid)
    Controls locality vs. variance; theory requires log^7(nd)/G→0 and G≤Δ/2; practice uses {60,80,100} or multiscale aggregation.
  • minimum-length η
    Filters short threshold exceedances in initial estimation; default 0.15, claimed stable in [0.1,0.3].
  • trimming ρ
    Excludes transition region when estimating local signal direction; fixed at 0.1.
  • active-set threshold multiplier c (or w+)
    Data-driven via separation-ratio maximization over a grid; theory only gives sufficient conditions involving unknown population quantities.
  • bootstrap replications B and level α
    B=200, α=0.05 used throughout numerics; asymptotic validity as B→∞.
axioms (4)
  • domain assumption Observations X1,...,Xn are independent (model (1.1) and all Hoeffding decompositions).
    Used for covariance calculations of the moving-window process and for the Gaussian approximation; mild AR(1) already degrades performance in Appendix H.
  • domain assumption Kernel h is antisymmetric and the first-order Hoeffding projections are non-degenerate with uniform sub-exponential or (2+ℓ)-moments (Assumptions A.1–A.3, B.2–B.3).
    Required for residual negligibility, Gaussian approximation, and bootstrap validity; verified for linear, sign, and variance kernels in Appendix D.
  • domain assumption Minimum spacing Δ≥2G and signal strength √G·θ♢ ≳ √log(nd) (Assumptions B.1, C.1).
    Ensures at most one change-point per window and that the deterministic signal dominates the max-norm fluctuation.
  • standard math High-dimensional Gaussian approximation and multiplier-bootstrap results of Chernozhukov et al. (2013,2017) extend to the moving-window U-statistic process.
    Core technical step in the proof of Theorem 3.1; the paper supplies the necessary uniform residual and covariance-approximation arguments.
invented entities (1)
  • U-PRA (U-statistic Projection Refinement Algorithm) independent evidence
    purpose: Refines initial ℓ∞ locations by projecting onto an estimated active set of coordinates, achieving the optimal localization rate and enabling the limiting distribution.
    Algorithmic construction; independent evidence is the theoretical rate and the simulation improvement over the initial estimator.

reviewed 2026-07-14 · how reviews work

0 comments
Cite this review

Pith. "Pith review of A General U-Statistic Framework for High-Dimensional Multiple Change-Point Analysis." pith.science (2026). https://pith.science/paper/QARTWTTO

@misc{pith2026260711256,
  author       = {Pith},
  title        = {Pith review of: A General U-Statistic Framework for High-Dimensional Multiple Change-Point Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QARTWTTO}},
  note         = {Machine review of arXiv:2607.11256}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

High-dimensional change-point analysis is essential in modern statistical inference. However, existing methods are often designed either for specific parameters (e.g., mean or variance) or for particular tasks (e.g., testing or estimation), making them difficult to generalize. Moreover, they typically rely on restrictive distributional assumptions, limiting their robustness to heavy-tailed data. We propose a unified framework for testing, estimating, and inferring multiple change points in high-dimensional data. Our approach leverages a two-sample U-statistic within a moving window, allowing flexible kernel function selection to accommodate structural changes in general parameters such as variance changes or robust statistics. For testing, we develop an L-infinity norm-based statistic with a high-dimensional multiplier bootstrap procedure, achieving minimax-optimal power under sparse alternatives. For estimation, we construct an initial estimator for the change-point number and locations and refine it using the U-statistic Projection Refinement Algorithm (U-PRA), attaining minimax-optimal localization rates. We further derive the asymptotic distribution of refined estimators, enabling valid confidence interval construction. Extensive numerical experiments demonstrate the better performance of our method across various settings, including heavy-tailed distributions. Applications to genomic copy number variation data highlight its practical utility. An R package implementing the proposed method, U-PRA, is publicly available at https://github.com/liubin0145/R-codes-UPRA/.

Figures

Figures reproduced from arXiv: 2607.11256 by Bin Liu, Yufeng Liu.

Figure 1
Figure 1. Figure 1: Local moving-window configuration around a change point. The [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of the moving-window statistic and the initial estimation [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Local neighborhoods used in the projection-based refinement step. [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Empirical powers under with different distributions. [PITH_FULL_IMAGE:figures/full_fig_p026_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Hausdorff localization errors under different error distributions and band [PITH_FULL_IMAGE:figures/full_fig_p030_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Segmented arrayCGH profiles using the linear and sign kernels. The [PITH_FULL_IMAGE:figures/full_fig_p036_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Residual autocorrelation after removing segment-wise means. The box [PITH_FULL_IMAGE:figures/full_fig_p037_7.png] view at source ↗

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

232 extracted references · 15 linked inside Pith

  1. [1]

    Electronic Journal of Statistics , volume=

    A robust bootstrap change point test for high-dimensional location parameter , author=. Electronic Journal of Statistics , volume=. 2022 , publisher=

    2022

  2. [2]

    Nature genetics , volume=

    Regional copy number--independent deregulation of transcription in cancer , author=. Nature genetics , volume=. 2006 , publisher=

    2006

  3. [3]

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

    Finite sample change point inference and identification for high-dimensional mean vectors , author =. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume =

  4. [4]

    Jackknife multiplier bootstrap: finite sample approximations to the

    Chen, Xiaohui and Kato, Kengo , journal =. Jackknife multiplier bootstrap: finite sample approximations to the. 2020 , doi =

    2020

  5. [5]

    Journal of Multivariate Analysis , volume =

    High-dimensional data analysis: Change point detection via bootstrap MOSUM , author =. Journal of Multivariate Analysis , volume =. 2025 , doi =

    2025

  6. [6]

    and Owens, D

    Cho, H. and Owens, D. , title =. Electronic Journal of Statistics , year =

  7. [7]

    and Zhu, H

    Zhou, H. and Zhu, H. and Wang, X. , title =. Journal of Multivariate Analysis , year =

  8. [8]

    Journal of the American Statistical Association , volume=

    Simultaneous inference for high-dimensional linear models , author=. Journal of the American Statistical Association , volume=. 2017 , publisher=

    2017

  9. [9]

    Test , volume=

    High-dimensional simultaneous inference with the bootstrap , author=. Test , volume=. 2017 , publisher=

    2017

  10. [10]

    Statistica Sinica , pages=

    A study of error variance estimation in lasso regression , author=. Statistica Sinica , pages=. 2016 , publisher=

    2016

  11. [12]

    arXiv preprint arXiv:2209.08892 , year=

    High-dimensional data segmentation in regression settings permitting heavy tails and temporal dependence , author=. arXiv preprint arXiv:2209.08892 , year=

    Pith/arXiv arXiv

  12. [13]

    2021 , journal =

    Optimal Covariance Change Point Localization in High Dimensions , author =. 2021 , journal =

    2021

  13. [14]

    arXiv preprint arXiv:2308.04368 , year=

    Multiple Testing of Local Extrema for Detection of Structural Breaks in Piecewise Linear Models , author=. arXiv preprint arXiv:2308.04368 , year=

    arXiv

  14. [15]

    Statistica Sinica , pages=

    A unified framework for change point detection in high-dimensional linear models , author=. Statistica Sinica , pages=

  15. [16]

    2021 , journal=

    Inference on the change point under a high dimensional sparse mean shift , author=. 2021 , journal=

    2021

  16. [17]

    arXiv preprint arXiv:2207.12453 , year=

    Change point inference in high-dimensional regression models under temporal dependence , author=. arXiv preprint arXiv:2207.12453 , year=

    Pith/arXiv arXiv

  17. [18]

    International Conference on Artificial Intelligence and Statistics , pages=

    Denoising and change point localisation in piecewise-constant high-dimensional regression coefficients , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2022 , organization=

    2022

  18. [19]

    Journal of Royal Statistical Society, Series B , volume=

    Confidence Intervals and Regions for the LASSO using Stochastic Variational Inequality Techniques in Optimization , author=. Journal of Royal Statistical Society, Series B , volume=

  19. [20]

    Journal of the American Statistical Association , volume=

    Confidence Intervals for Sparse Penalized Regression , author=. Journal of the American Statistical Association , volume=

  20. [21]

    The Annals of Statistics , volume=

    On asymptotically optimal confidence regions and tests for high-dimensional models , author=. The Annals of Statistics , volume=. 2014 , publisher=

    2014

  21. [22]

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

    Confidence intervals for low dimensional parameters in high dimensional linear models , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2014 , publisher=

    2014

  22. [23]

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

    Regression shrinkage and selection via the lasso , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 1996 , publisher=

    1996

  23. [24]

    High-dimensional graphs and variable selection with the lasso , author=. The. 2006 , publisher=

    2006

  24. [25]

    The Annals of Statistics , volume=

    Uniform change point tests in high dimension , author=. The Annals of Statistics , volume=

  25. [26]

    Journal of

    Limit theorems for the union-intersection test , author=. Journal of. 1995 , publisher=

    1995

  26. [27]

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

    The lasso for high dimensional regression with a possible change point , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2016 , publisher=

    2016

  27. [28]

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

    Sure independence screening for ultrahigh dimensional feature space , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2008 , publisher=

    2008

  28. [29]

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

    Robustness and accuracy of methods for high dimensional data analysis based on Student's t-statistic , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2011 , publisher=

    2011

  29. [30]

    Preprint arXiv:1808.02648 , year=

    A unified framework for testing high dimensional parameters: a data-adaptive approach , author=. Preprint arXiv:1808.02648 , year=

    Pith/arXiv arXiv

  30. [31]

    On the Maximal Perimeter of a Convex Set in

    Nazarov, Fedor , journal=. On the Maximal Perimeter of a Convex Set in. 2003 , publisher=

    2003

  31. [32]

    Electronic Journal of Statistics , volume=

    Relevant change points in high dimensional time series , author=. Electronic Journal of Statistics , volume=. 2018 , publisher=

    2018

  32. [33]

    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

  33. [34]

    Concentration

    Boucheron, St. Concentration. 2013 , publisher=

    2013

  34. [35]

    Submitted , year=

    Simultaneous Change Point Detection and Structure Recovery for High Dimensional Gaussian Graphical Models , author=. Submitted , year=

  35. [36]

    The Annals of Statistics , volume=

    Simultaneous analysis of Lasso and Dantzig selector , author=. The Annals of Statistics , volume=. 2009 , publisher=

    2009

  36. [37]

    Journal of Royal Statistical Society, Series B (Statistical Methodology) , pages=

    A Unified Data-adaptive Framework for High Dimensional Change Point Detection , author=. Journal of Royal Statistical Society, Series B (Statistical Methodology) , pages=

  37. [38]

    The Annals of Statistics , volume=

    High-dimensional regression with noisy and missing data: Provable guarantees with non-convexity , author=. The Annals of Statistics , volume=

  38. [39]

    Statistica Sinica , volume=

    Two-sample tests for high-dimensional linear regression with an application to detecting interactions , author=. Statistica Sinica , volume=

  39. [40]

    The Annals of Probability , volume=

    Central limit theorems and bootstrap in high dimensions , author=. The Annals of Probability , volume=. 2017 , publisher=

    2017

  40. [41]

    Gaussian and bootstrap approximations for high-dimensional

    Chen, Xiaohui , journal=. Gaussian and bootstrap approximations for high-dimensional

  41. [42]

    The Annals of Statistics , volume=

    Gaussian approximations and multiplier bootstrap for maxima of sums of high-dimensional random vectors , author=. The Annals of Statistics , volume=

  42. [43]

    The Annals of Statistics , volume=

    The Dantzig selector: statistical estimation when p is much larger than n , author=. The Annals of Statistics , volume=

  43. [44]

    Annals of Statistics , volume=

    High-dimensional graphs and variable selection with the Lasso , author=. Annals of Statistics , volume=

  44. [45]

    The Journal of Machine Learning Research , volume=

    On Model Selection Consistency of Lasso , author=. The Journal of Machine Learning Research , volume=

  45. [46]

    Journal of the American Statistical Association , volume=

    Simultaneous Inference for High-dimensional Linear Models , author=. Journal of the American Statistical Association , volume=

  46. [47]

    The Journal of Machine Learning Research , volume=

    Confidence intervals and hypothesis testing for high-dimensional regression , author=. The Journal of Machine Learning Research , volume=

  47. [48]

    2011 , publisher=

    Statistics for high-dimensional data: methods, theory and applications , author=. 2011 , publisher=

    2011

  48. [49]

    Biometrika , volume=

    Control charts with warning lines , author=. Biometrika , volume=. 1955 , publisher=

    1955

  49. [50]

    Journal of the American Statistical Association , volume=

    TESTS OF THE HYPOTHESIS THAT A LINEAR REGRESSION SYSTEM OBEYS TWO SEPARATE REGIMES , author=. Journal of the American Statistical Association , volume=

  50. [51]

    Journal of the American Statistical Association , volume=

    The estimation of the parameters of a linear regression system obeying two separate regimes , author=. Journal of the American Statistical Association , volume=

  51. [52]

    Statistics , volume=

    Detecting Changes in Linear Regressions , author=. Statistics , volume=

  52. [53]

    Journal of the American Statistical Association , volume=

    A New Approach to Estimating Switching Regressions , author=. Journal of the American Statistical Association , volume=

  53. [54]

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

    Techniques for Testing the Constancy of Regression Relationships Over Time , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=

  54. [55]

    Journal of Multivariate Analysis , volume=

    Limit theorems for change in linear regression , author=. Journal of Multivariate Analysis , volume=

  55. [56]

    Econometrica , volume=

    Estimating and testing linear models with multiple structural changes , author=. Econometrica , volume=

  56. [57]

    1997 , publisher=

    Limit theorems in change-point analysis , author=. 1997 , publisher=

    1997

  57. [58]

    Journal of the American Statistical Association , volume=

    Oracle Estimation of a Change Point in High Dimensional Quantile Regression , author=. Journal of the American Statistical Association , volume=

  58. [60]

    In: 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton) , publisher=

    Change-point estimation in high dimensional linear regression models via sparse group lasso , author=. In: 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton) , publisher=

  59. [61]

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

    Goodness-of-fit testing in high-dimensional generalized linear models , author=. Journal of Royal Statistical Society, Series B (Statistical Methodology) , volume=

  60. [62]

    Soviet Math

    Detecting disorder in multidimensional random process , author=. Soviet Math. Dokl , volume=

  61. [63]

    Journal of the American Statistical Association , volume=

    Objective Criteria for the Evaluation of Clustering Methods , author=. Journal of the American Statistical Association , volume=

  62. [64]

    Quarterly Journal of Economics , volume=

    Tipping and the Dynamics of Segregation , author=. Quarterly Journal of Economics , volume=

  63. [65]

    Journal of Economic Dynamics and Control , volume=

    Econometric Issues in the Analysis of Contagion , author=. Journal of Economic Dynamics and Control , volume=

  64. [66]

    Statistics and Computing , volume=

    A novel and fast methodology for simultaneous multiple structural break estimation and variable selection for nonstationary time series models , author=. Statistics and Computing , volume=

  65. [67]

    Technometrics , volume=

    Most Recent Changepoint Detection in Panel Data , author=. Technometrics , volume=

  66. [68]

    Journal of the American Statistical Association , volume=

    Structural Break Estimation for Nonstationary Time Series Models , author=. Journal of the American Statistical Association , volume=

  67. [69]

    Statistics and Computing , volume=

    On optimal multiple changepoint algorithms for large data , author=. Statistics and Computing , volume=

  68. [70]

    Journal of the American Statistical Association , volume=

    Group LASSO for Structural Break Time Series , author=. Journal of the American Statistical Association , volume=

  69. [71]

    Bernoulli , volume=

    Piecewise quantile autoregressive modeling for nonstationary time series , author=. Bernoulli , volume=

  70. [72]

    Annals of Statistics , volume=

    Weak Signal Identification and Inference in Penalized Model Selection , author=. Annals of Statistics , volume=. 2016 , pages=

    2016

  71. [73]

    Statistica Sinica , volume=

    FULLY EFFICIENT ROBUST ESTIMATION, OUTLIER DETECTION AND VARIABLE SELECTION VIA PENALIZED REGRESSION , author=. Statistica Sinica , volume=. 2018 , pages=

    2018

  72. [74]

    and Folstein, Susan E

    Folstein, Marshal F. and Folstein, Susan E. and Mchugh, Paul R. , journal=. "

  73. [75]

    Journal of the American Statistical Association , year=

    Sparse Regression Incorporating Graphical Structure Among Predictors , author=. Journal of the American Statistical Association , year=

  74. [76]

    Multi-Modal Multi-Task Learning for Joint Prediction of Multiple Regression and Classification Variables in

    Zhang, Daoqiang and Shen, Dinggang , journal=. Multi-Modal Multi-Task Learning for Joint Prediction of Multiple Regression and Classification Variables in. 2012 , volume=

    2012

  75. [77]

    The Annals of Statistics , volume=

    Minimax rates in sparse, high-dimensional change point detection , author=. The Annals of Statistics , volume=. 2021 , publisher=

    2021

  76. [78]

    Structural Break Detection in High-Dimensional Non-Stationary

    Safikhani, Abolfazl and Shojaie, Ali , journal=. Structural Break Detection in High-Dimensional Non-Stationary. 2022 , volume=

    2022

  77. [79]

    Biometrics , year=

    Drawing inferences for high‐dimensional linear models: A selection‐assisted partial regression and smoothing approach , author=. Biometrics , year=

  78. [80]

    Lin and Rebecca Willett , title =

    Daren Wang and Zifeng Zhao and Kevin Z. Lin and Rebecca Willett , title =. The Journal of Machine Learning Research , year =

  79. [81]

    Restricted eigenvalue properties for correlated

    Raskutti, Garvesh and Wainwright, Martin J and Yu, Bin , journal=. Restricted eigenvalue properties for correlated. 2010 , publisher=

    2010

  80. [82]

    Electronic Journal of Statistics , volume=

    On the conditions used to prove oracle results for the lasso , author=. Electronic Journal of Statistics , volume=. 2009 , publisher=

    2009

Showing first 80 references.

This paper was first reviewed by grok-4.5 on July 14, 2026.