Pith. sign in

REVIEW 2 major objections 4 minor 47 references

Change-Points Detection and Support Recovery for Spatially Indexed Functional Data

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Under weak separability, a single FPCA pipeline detects the global change-point and recovers the affected spatial sites with FDR control.

desk verdict A solid integration of weak separability, kernel CUSUM, and SDA for spatial functional change-points, but the FDR proof currently rests on an arithmetically false variance formula that must be fixed. read the letter →

arxiv 2506.07206 v2 pith:ZMSVRLPA submitted 2025-06-08 stat.ME

Fengyi Song , Decai Liang , Changliang Zou This is my paper
classification stat.ME
keywords change-pointdetectionfunctionaldataspatiotemporalweakseparabilityfalsediscoveryratespatialmultipletestingprincipalcomponentanalysisCUSUM
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that two previously separate tasks for spatially indexed functional data—detecting the year when the mean pattern changes and identifying which spatial locations actually shifted—can be solved in one framework. The enabling assumption is weak separability: every location's curve is built from the same temporal basis functions, with uncorrelated scores across components, while spatial dependence within each component can be arbitrary. Under that assumption, a kernel-weighted CUSUM statistic detects the global break with a null distribution that is a quadratic form of Brownian bridges, and the estimated change-point is within $O_p(1)$ of the true one. A functional version of symmetrized data aggregation then controls the false discovery proportion at $\alpha+o_p(1)$ without needing pointwise $p$-values. Applied to Eastern China precipitation, the pipeline finds a break in 1988 and maps the stations that drove it.

What carries the argument

The load-bearing object is weak separability, the representation $X_i(s;t)=\mu_i(s,t)+\sum_{r\ge1}\xi_{ir}(s)\psi_r(t)$ in which the FPC scores are uncorrelated across $r$; it makes the marginal covariance's eigenfunctions the common temporal basis and reduces each component's spatial dependence to a matrix $\Sigma_r$. On top of it sit three mechanisms. A differencing covariance estimator $\hat{H}(t,t')$ from adjacent observations keeps the estimate consistent even when a mean break is present, so the estimated eigenfunctions $\hat\psi_r$ and scores $\hat\eta_{\tau,r}(s_j)$ are reliable. The kernel CUSUM $Q_h(\tau)=\sum_{r=1}^R \hat{\boldsymbol\eta}_{\tau,r}^\top K_h \hat{\boldsymbol\eta}_{\tau,r}$ aggregates squared projected CUSUMs across neighboring locations, which is what turns spatial clustering into detection power. Finally, the functional SDA statistic $W_j=\sum_{r=1}^R \hat\eta^{O}_{\hat\tau,r}(s_j)\,\tilde\eta^{E}_{\hat\tau,r}(s_j)$ combines an odd-index split with a kernel-weighted even-index split; its asymptotic sign symmetry under the null is the mechanism that makes a data-driven threshold control FDP without pointwise $p$-values.

What would settle it

Simulate data from a process with region-specific temporal bases (for example, different $\psi_r$ in the northern and southern halves of the spatial domain) and run Algorithm 1: if the empirical null distribution of $Q_h$ does not match the quadratic-form-of-Brownian-bridges limit or the fSDA FDP exceeds $\alpha$, weak separability rather than the detection machinery is doing the work.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that weak separability gives a valid foundation for both global and local inference under a single mean change. The authors define a spatiotemporal process as weakly separable when $X_i(s;t)=\mu_i(s,t)+\sum_{r\ge1}\xi_{ir}(s)\psi_r(t)$ with mutually uncorrelated FPC scores across $r$; the temporal basis is common across space, while each component has its own spatial covariance. Given this structure, the differencing-based FPCA estimators are consistent under both null and alternative, and the kernel CUSUM $Q_h(\tau)$ converges to a quadratic form $\sum_{r=1}^R \mathbb{B}_r^\top(x) K_h \mathbb{B}_r(x)$ of Brownian bridges with spatial correlation $\mathbb{P}_r$. The same machinery yields the change-point estimator $\hat\tau$ with $|\hat\tau-\tau^*|=O_p(1)$ and, after an order-preserved sample split, ranking statistics $W_j$ whose sign symmetry under the null lets the functional SDA rule control FDP at $\alpha+o_p(1)$. The authors also claim full recovery of identifiable effect locations with probability tending to one.

Load-bearing premise

The entire construction depends on weak separability—every spatial location must share the same temporal basis functions with uncorrelated scores across components; if different regions have genuinely different temporal structures, the FPCA-based statistics lose their stated justification.

Editorial extensions

If this is right

  • A single common temporal basis is enough to justify both change-point testing and support recovery, so the method avoids the more restrictive space-time separability used by earlier spatial-functional change tests.
  • The kernel aggregation improves power when locations are dense or the changed region is spatially clustered, and the improvement grows with the number of locations $p$.
  • The change-point estimator's $O_p(1)$ error means the detected break year is consistent in a strong sense: the estimated and true indices differ by a bounded random amount, not just at the slower rate $\hat\tau/n\to\theta_0$.
  • The support recovery step controls false discoveries at level $\alpha$ asymptotically without modeling $p$-values, so it remains valid under dependence structures where two-component mixture models are questionable.
  • If the mean shift is orthogonal to the span of the retained eigenfunctions, the test has no power by the paper's own divergence condition; the FVE-based truncation must keep the signal components.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: because the FDP argument only needs sign symmetry of the ranking statistics, the same sample-splitting threshold could be applied to other spatially indexed functional summaries, such as local covariance or quantile shifts, whenever a matching symmetry can be established; the paper does not claim this extension.
  • Inference: the bandwidth rule based on the fitted correlation range is heuristic; a natural testable extension is to choose $h$ by minimizing an estimate of the FDP or by cross-validating power, which the paper does not explore.
  • Inference: applying the pipeline to the 1961–2013 Eastern China precipitation records yields 1988 as a break year, consistent with broader climate records; one could test whether the recovered station set reproduces under different FVE levels or on neighboring regions, which the paper only partially addresses.
  • Inference: the one-change-point focus means sequences with multiple breaks would need iterative application; the paper says additional change-points can be detected as needed but does not analyze error propagation under repeated estimation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes a unified framework for detecting a single change-point in spatially indexed functional data and for recovering the spatial support of the change. It models the spatiotemporal covariance as weakly separable, estimates the common temporal eigenbasis by a differencing covariance estimator, builds a kernel-weighted CUSUM statistic Q_h, derives its null distribution as a quadratic form of Brownian bridges, and estimates the change-point by maximizing Q_h. For support recovery, it introduces a functional symmetrized data aggregation (fSDA) procedure based on sample splitting and kernel-weighted ranking statistics, and proves asymptotic FDR control. Simulations and a precipitation data application illustrate the methodology.

Significance. If the proofs are completed, the paper makes a useful contribution: it relaxes the strong separability assumption used by Gromenko, Kokoszka and Reimherr (2017), gives explicit null distributions for both max- and sum-type kernel CUSUM statistics, and extends the p-value-free SDA idea to functional spatial data. The differencing estimator for the marginal covariance is a genuine technical device that remains consistent under the alternative, and the proof structure, based on oracle statistics plus uniform estimation-error bounds, is coherent. The paper also includes a real-data application with separability diagnostics. However, an arithmetic error in a key lemma and a mismatch between an assumption and the stated moderate-deviation range currently undermine the written proof of the FDR guarantee.

major comments (2)
  1. [Appendix A, Lemma A.2(a); Eq. (5)] The claimed variance of eta*_{j,r} is arithmetically false. Direct calculation with standardized scores xi*_{ir,j} gives var(eta*_{j,r}) = [tau*(n-tau*)/(n^{3/2})]^2 [1/tau* + 1/(n-tau*)] = tau*(n-tau*)/n^2 = theta0(1-theta0), not theta0(1-theta0)(1-2theta0). The factor (1-2theta0) is negative for theta0>1/2, and the statement that theta0<1/2 without loss of generality does not repair the algebra. This constant is used in Lemma C.3 to standardize eta*^T tilde{eta}* and in Lemma C.4 to evaluate the conditional variance and covariance of W*_j, and Lemma C.4 is the route to the symmetry property (14) that Theorem 4.1 relies on. The theorem is probably salvageable because the symmetry of the centered conditional distribution does not depend on the exact variance, and replacing the wrong constant by theta0(1-theta0) preserves the standardization and the correlation bound in Lemma C.4. But as written the proof of Theorem 4.1 is internally inconsistent; the authors should correct Lemma A.2(a), the Section 4.2 statement that T^O_{j,r} is asymptotically N(0,1) (at the scaling in (5) it is N(0,theta0(1-theta0))), and the affected displays in Lemmas C.3-C.4.
  2. [Section 4.3, Assumption 4.1 and Lemma C.2] There is a second gap in the FDR proof: Lemma C.2, the moderate-deviation result for two-sample t-statistics, is stated to hold uniformly for x in (0, o((n1+n2)^{1/6})), while Lemma A.3 establishes that the threshold t* in Lemma C.3 satisfies t* <= C_eta log p. Assumption 4.1 only gives p = o(exp(n^{1/4})), i.e. log p = o(n^{1/4}), which does not imply log p = o(n^{1/6}). Thus the range of thresholds t in [0,t*] is not guaranteed to fall inside the range where Lemma C.2 applies. Since Lemma C.3's uniform symmetry of G(t)/G_-(t) is the basis for the empirical symmetry property (14), the proof of Theorem 4.1 has a second load-bearing gap. This is fixable by strengthening Assumption 4.1 to p = o(exp(n^{1/6})) or by replacing Lemma C.2 with a sub-Gaussian tail result that controls the symmetry ratio over the larger range, but the manuscript should state and prove the correct condition.
minor comments (4)
  1. [Section 3.2, after Theorem 3.2] The statement that n^{-1} Q_h^{max} = O_p(1) is weaker than the actual result in (32), which gives convergence to sup_{0<x<1} g(x); the authors should state the limit directly to avoid understating the theorem.
  2. [Tables 3 and 7] The method labels 'Q0' and 'Qh' are typographically inconsistent with the notation Q_0 and Q_h used in the text; please unify the notation.
  3. [Algorithm 2 and Assumption 4.5] The bandwidth selection discussion before Algorithm 2 says that a relatively large threshold rho is determined for practical applicability, but the formal guarantee in Theorem 4.1 is tied to Assumption 4.5; the text should make explicit how the practical choice of h is intended to satisfy or approximate that assumption.
  4. [Section 1.3 and Assumptions 4.2-4.5] The paper describes the method as 'model-free in several aspects', but the FDR theory relies on weak separability, sub-Gaussian scores, the strong-correlation sparsity condition, and the bandwidth condition; a more measured wording would better reflect the scope of the theoretical guarantees.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: null distribution, change-point estimation, and FDR control are derived analytically; self-citations are not load-bearing.

full rationale

After walking the derivation chain, I find no circular step. The three main claims are derived analytically, not fitted. Theorem 3.1's null distribution follows from a functional CLT on FPC-score partial sums (Lemma A.1) plus Slutsky/continuous-mapping arguments; the limiting Brownian-bridge quadratic form is not used to define Q_h. The change-point estimator \hat\tau is the argmax of Q_h(\tau), and Corollary 3.1 derives its O_p(1) localization from the uniform convergence in (32) and positivity of g_R^*; no parameter is fitted to the target. The fSDA threshold L is constructed from the empirical distribution of W_j via (13)/(B.5), and Theorem 4.1 proves FDP≤α+o_p(1) using the symmetry property (14), which is established in Lemmas A.3 and C.3–C.5 from independence of the sample splits, moderate-deviation bounds, and Assumptions 4.1–4.5; the threshold is not chosen to match the observed FDP. The self-citations (Liang et al. 2023 for weak separability; Du et al. 2021, Chen et al. 2023, and Luo et al. 2025, involving co-authors Zou and Liang) are used as external building blocks or assumption checks, and the paper supplies its own lemmas and proofs, so the load-bearing arguments do not reduce to an unverified self-citation. The apparent arithmetic inconsistency in Lemma A.2(a), which claims variance θ0(1−θ0)(1−2θ0) versus the direct calculation θ0(1−θ0), is a correctness concern in the proof of the FDR theorem, not a circularity: it does not make the conclusion an input of the construction. It should be fixed, but it does not affect the circularity verdict.

Assumptions & free parameters 2 free parameters · 7 assumptions · 0 invented entities

The method's correctness rests primarily on the weak separability assumption and the dependence and bandwidth conditions in Section 4. These are stated explicitly and partially tested on the real data. The tuning parameters R and h are selected by heuristic rules.

free parameters (2)
  • bandwidth h = 487.1 km in application; simulation uses ρ=0.05 yielding h∈[0.18,0.47]
    Spatial kernel bandwidth controlling the neighborhood aggregation in Q_h and fSDA; selected via a threshold on the estimated spatial correlation function, not part of the model.
  • truncation number R = 5 in application (FVE=90%)
    Number of functional principal components retained; chosen by fraction of variance explained, affects power but not asymptotic validity.
assumptions (7)
  • domain assumption Weak separability: there exists a common orthonormal temporal basis such that FPC scores are uncorrelated across components for all spatial locations.
    Definition 1 is the core structural assumption enabling the marginal covariance eigenbasis to serve as the temporal FPC basis. Tested on the real data in Appendix E.2.
  • domain assumption The replications ε_i are i.i.d. across i with continuous covariance functions.
    Assumption 2.2 enables replication-based FPCA and the functional CLT for the CUSUM process.
  • domain assumption Finite fourth moments and distinct eigenvalues with bounded spectral norms.
    Assumptions 2.3 and 2.4 ensure consistency and eigen-estimation rates for the FPCA estimators.
  • domain assumption Standardized FPC scores are sub-Gaussian with a common variance proxy, and the sum of eigenvalues across components is bounded.
    Assumption 4.2 provides the tail bounds and concentration inequalities needed for the uniform convergence of the ranking statistics W_j.
  • domain assumption The number of locations with strong spatial correlation is bounded by l_p with l_p/b_p converging to zero.
    Assumption 4.4 is the dependence condition required for the uniform symmetry property (14) and hence for FDR control.
  • domain assumption The bandwidth h satisfies sup over neighbors of sqrt(n)||δ_k - δ_j|| = O(sqrt(log p)), approximately nh^2 = O(log p).
    Assumption 4.5 ensures that kernel aggregation does not distort the signal, preserving the symmetry of the null statistics.
  • domain assumption The number of identifiable alternatives b_p diverges to infinity with p.
    Assumption 4.3 is a standard condition for FDP control, ensuring enough signal for the threshold to be informative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Change-Points Detection and Support Recovery for Spatially Indexed Functional Data." pith.science (2026). https://pith.science/paper/ZMSVRLPA

@misc{pith2026250607206,
  author       = {Pith},
  title        = {Pith review of: Change-Points Detection and Support Recovery for Spatially Indexed Functional Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZMSVRLPA}},
  note         = {Machine review of arXiv:2506.07206}
}
read the original abstract

Large volumes of spatiotemporal data, characterized by high spatial and temporal variability, may experience structural changes over time. Unlike traditional change-point problems, each sequence in this context consists of function-valued curves observed at multiple spatial locations, with typically only a small subset of locations affected. This paper addresses two key issues: detecting the global change-point and identifying the spatial support set, within a unified framework tailored to spatially indexed functional data. By leveraging a weakly separable cross-covariance structure -- an extension beyond the restrictive assumption of space-time separability -- we incorporate functional principal component analysis into the change-detection methodology, while preserving common temporal features across locations. A kernel-based test statistic is further developed to integrate spatial clustering pattern into the detection process, and its local variant, combined with the estimated change-point, is employed to identify the subset of locations contributing to the mean shifts. To control the false discovery rate in multiple testing, we introduce a functional symmetrized data aggregation approach that does not rely on pointwise p-values and effectively pools spatial information. We establish the asymptotic validity of the proposed change detection and support recovery method under mild regularity conditions. The efficacy of our approach is demonstrated through simulations, with its practical usefulness illustrated in an application to China's precipitation data.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 46 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" url if FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if ne...

  2. [2]

    barticle [author] Aston , John AD J. A. Kirch , Claudia C. ( 2012 ). Detecting and estimating changes in dependent functional data . Journal of Multivariate Analysis 109 204--220 . barticle

    work page 2012

  3. [3]

    barticle [author] Aston , John AD J. A. , Pigoli , Davide D. , Tavakoli , Shahin S. et al. ( 2017 ). Tests for separability in nonparametric covariance operators of random surfaces . The Annals of Statistics 45 1431--1461 . barticle

    work page 2017

  4. [4]

    , Rice , Gregory G

    barticle [author] Aue , Alexander A. , Rice , Gregory G. S \"o nmez , Ozan O. ( 2018 ). Detecting and dating structural breaks in functional data without dimension reduction . Journal of the Royal Statistical Society Series B: Statistical Methodology 80 509--529 . barticle

    work page 2018

  5. [5]

    barticle [author] Barber , Rina Foygel R. F. Cand \`e s , Emmanuel J. E. J. ( 2015 ). Controlling the false discovery rate via knockoffs . The Annals of Statistics 43 2055--2085 . barticle

    work page 2015

  6. [6]

    Heller , Ruth R

    barticle [author] Benjamini , Yoav Y. Heller , Ruth R. ( 2007 ). False discovery rates for spatial signals . Journal of the American Statistical Association 102 1272--1281 . barticle

    work page 2007

  7. [7]

    Hochberg , Yosef Y

    barticle [author] Benjamini , Yoav Y. Hochberg , Yosef Y. ( 1995 ). Controlling the false discovery rate: a practical and powerful approach to multiple testing . Journal of the Royal Statistical Society Series B: Statistical Methodology 57 289--300 . barticle

    work page 1995

  8. [8]

    Yekutieli , Daniel D

    barticle [author] Benjamini , Yoav Y. Yekutieli , Daniel D. ( 2001 ). The control of the false discovery rate in multiple testing under dependency . Annals of statistics 1165--1188 . barticle

    work page 2001

Show all 47 references
  1. [9]

    , Gabrys , Robertas R

    barticle [author] Berkes , Istv \'a n I. , Gabrys , Robertas R. , Horv \'a th , Lajos L. Kokoszka , Piotr P. ( 2009 ). Detecting changes in the mean of functional observations . Journal of the Royal Statistical Society Series B: Statistical Methodology 71 927--946 . barticle

    2009

  2. [10]

    ( 2000 )

    bbook [author] Bosq , Denis D. ( 2000 ). Linear processes in function spaces: theory and applications . Springer Science & Business Media . bbook

    2000

  3. [11]

    barticle [author] Brown , D Andrew D. A. , Lazar , Nicole A N. A. , Datta , Gauri S G. S. , Jang , Woncheol W. McDowell , Jennifer E J. E. ( 2014 ). Incorporating spatial dependence into Bayesian multiple testing of statistical parametric maps in functional neuroimaging . Neur...

    2014

  4. [12]

    barticle [author] Brunsell , N. A. N. A. ( 2010 ). A multiscale information theory approach to assess spatial-temporal variability of daily precipitation . Journal of Hydrology 385 165--172 . barticle

    2010

  5. [13]

    barticle [author] Cai , T. Tony T. T. Liu , Weidong W. ( 2016 ). Large-scale multiple testing of correlations . Journal of the American Statistical Association 111 229--240 . barticle

    2016

  6. [14]

    barticle [author] Cai , T Tony T. T. , Sun , Wenguang W. Xia , Yin Y. ( 2022 ). LAWS: A locally adaptive weighting and screening approach to spatial multiple testing . Journal of the American Statistical Association 117 1370--1383 . barticle

    2022

  7. [15]

    ( 2007 )

    barticle [author] Cao , Hongyuan H. ( 2007 ). Moderate deviations for two sample t -statistics . ESAIM: Probability and Statistics 11 264--271 . barticle

    2007

  8. [16]

    Wu , WeiBiao W

    barticle [author] Cao , Hongyuan H. Wu , WeiBiao W. ( 2015 ). Changepoint estimation: another look at multiple testing problems . Biometrika 102 974--980 . barticle

    2015

  9. [17]

    , Ren , Haojie H

    barticle [author] Chen , Hui H. , Ren , Haojie H. , Yao , Fang F. Zou , Changliang C. ( 2023 ). Data-driven selection of the number of change-points via error rate control . Journal of the American Statistical Association 118 1415--1428 . barticle

    2023

  10. [18]

    Wikle , Christopher K C

    bbook [author] Cressie , Noel N. Wikle , Christopher K C. K. ( 2011 ). Statistics for spatio-temporal data . John Wiley & Sons . bbook

    2011

  11. [19]

    , Guo , Xu X

    barticle [author] Du , Lilun L. , Guo , Xu X. , Sun , Wenguang W. Zou , Changliang C. ( 2021 ). False discovery rate control under general dependence by symmetrized data aggregation . Journal of the American Statistical Association 118 607--621 . barticle

    2021

  12. [20]

    , Banerjee , Sudipto S

    barticle [author] Gao , Leiwen L. , Banerjee , Sudipto S. Ritz , Beate B. ( 2023 ). Spatial difference boundary detection for multiple outcomes using Bayesian disease mapping . Biostatistics 24 922--944 . barticle

    2023

  13. [21]

    , Kokoszka , Piotr P

    barticle [author] Gromenko , Oleksandr O. , Kokoszka , Piotr P. Reimherr , Matthew M. ( 2017 ). Detection of change in the spatiotemporal mean function . Journal of the Royal Statistical Society Series B: Statistical Methodology 79 29--50 . barticle

    2017

  14. [22]

    , Kokoszka , Piotr P

    barticle [author] H \"o rmann , Siegfried S. , Kokoszka , Piotr P. et al. ( 2010 ). Weakly dependent functional data . The Annals of Statistics 38 1845--1884 . barticle

    2010

  15. [23]

    Kokoszka , Piotr P

    bbook [author] Horv \'a th , Lajos L. Kokoszka , Piotr P. ( 2012 ). Inference for functional data with applications . Springer Science & Business Media . bbook

    2012

  16. [24]

    Eubank , Randall R

    bbook [author] Hsing , Tailen T. Eubank , Randall R. ( 2015 ). Theoretical foundations of functional data analysis, with an introduction to linear operators . John Wiley & Sons . bbook

    2015

  17. [25]

    Staicu , Ana-Maria A.-M

    barticle [author] Koner , Salil S. Staicu , Ana-Maria A.-M. ( 2023 ). Second-generation functional data . Annual review of statistics and its application 10 547--572 . barticle

    2023

  18. [26]

    , Li , Runze R

    barticle [author] Li , Degui D. , Li , Runze R. Shang , Han Lin H. L. ( 2024 ). Detection and estimation of structural breaks in high-dimensional functional time series . The Annals of Statistics 52 1716--1740 . barticle

    2024

  19. [27]

    , Zhang , Xianyang X

    barticle [author] Li , Bo B. , Zhang , Xianyang X. Smerdon , Jason E J. E. ( 2016 ). Comparison between spatio-temporal random processes and application to climate model data . Environmetrics 27 267--279 . barticle

    2016

  20. [28]

    , Huang , Hui H

    barticle [author] Liang , Decai D. , Huang , Hui H. , Guan , Yongtao Y. Yao , Fang F. ( 2023 ). Test of weak separability for spatially stationary functional field . Journal of the American Statistical Association 118 1606--1619 . barticle

    2023

  21. [29]

    ( 2013 )

    barticle [author] Liu , Weidong W. ( 2013 ). Gaussian graphical model estimation with false discovery rate control . The Annals of Statistics 41 2948--2978 . barticle

    2013

  22. [30]

    , Ray , Surajit S

    barticle [author] Liu , Chong C. , Ray , Surajit S. Hooker , Giles G. ( 2017 ). Functional principal component analysis of spatially correlated data . Statistics and Computing 27 1639--1654 . barticle

    2017

  23. [31]

    Shao , Qi-Man Q.-M

    barticle [author] Liu , Weidong W. Shao , Qi-Man Q.-M. ( 2014 ). Phase transition and regularized bootstrap in large-scale t -tests with false discovery rate control . The Annals of Statistics 42 2003--2025 . barticle

    2014

  24. [32]

    , Liu , Pengfei P

    barticle [author] Liu , Min M. , Liu , Pengfei P. , Guo , Ying Y. , Wang , Yanfang Y. , Geng , Xinxin X. , Nie , Zhenlong Z. Yu , Yang Y. ( 2019 ). Change-point analysis of precipitation and drought extremes in china over the past 50 years . Atmosphere 11 11 . barticle

    2019

  25. [33]

    , Zhang , Wei W

    barticle [author] Luo , Fangzhi F. , Zhang , Wei W. Liang , Decai D. ( 2025 ). Test of partial separability for multivariate functional data . Statistica Sinica published online . barticle

    2025

  26. [34]

    , Zhang , Qinghong Q

    barticle [author] Ng , Chan-Pang C.-P. , Zhang , Qinghong Q. Li , Wenhong W. ( 2021 ). Changes in hourly extreme precipitation over Eastern China from 1970 to 2019 dominated by synoptic-scale precipitation . Geophysical Research Letters 48 e2020GL090620 . barticle

    2021

  27. [35]

    barticle [author] Pacifico , Marco Perone M. P. , Genovese , Christopher C. , Verdinelli , Isabella I. Wasserman , Larry L. ( 2004 ). False discovery control for random fields . Journal of the American Statistical Association 99 1002--1014 . barticle

    2004

  28. [36]

    , Zou , Changliang C

    barticle [author] Paynabar , Kamran K. , Zou , Changliang C. Qiu , Peihua P. ( 2016 ). A change-point approach for phase-I analysis in multivariate profile monitoring and diagnosis . Technometrics 58 191--204 . barticle

    2016

  29. [37]

    , Zou , Changliang C

    barticle [author] Ren , Haojie H. , Zou , Changliang C. , Chen , Nan N. Li , Runze R. ( 2022 ). Large-scale datastreams surveillance via pattern-oriented-sampling . Journal of the American Statistical Association 117 794--808 . barticle

    2022

  30. [38]

    barticle [author] Risser , Mark D M. D. , Paciorek , Christopher J C. J. Stone , D \'a ith \' A D. A. ( 2019 ). Spatially dependent multiple testing under model misspecification, with application to detection of anthropogenic influence on extreme climate events . Journal of th...

    2019

  31. [39]

    , Reich , Brian J

    barticle [author] Sun , Wenguang W. , Reich , Brian J. B. J. , Tony Cai , T. T. , Guindani , Michele M. Schwartzman , Armin A. ( 2015 ). False discovery control in large-scale spatial multiple testing . Journal of the Royal Statistical Society Series B: Statistical Methodology...

    2015

  32. [40]

    , Chiou , Jeng-Min J.-M

    barticle [author] Wang , Jane-Ling J.-L. , Chiou , Jeng-Min J.-M. M \"u ller , Hans-Georg H.-G. ( 2016 ). Functional data analysis . Annual Review of Statistics and its application 3 257--295 . barticle

    2016

  33. [41]

    Feng , Long L

    barticle [author] Wang , Guanghui G. Feng , Long L. ( 2023 ). Computationally efficient and data-adaptive changepoint inference in high dimension . Journal of the Royal Statistical Society Series B: Statistical Methodology 85 936--958 . barticle

    2023

  34. [42]

    , Paynabar , Kamran K

    barticle [author] Yan , Hao H. , Paynabar , Kamran K. Shi , Jianjun J. ( 2018 ). Real-time monitoring of high-dimensional functional data streams via spatio-temporal smooth sparse decomposition . Technometrics 60 181--197 . barticle

    2018

  35. [43]

    barticle [author] Yun , Zhang Xianyang Z. X. Sooin Bo , Li L. ( 2022 ). Detection of local differences in spatial characteristics between two spatiotemporal random fields . Journal of the American Statistical Association 117 291--306 . barticle

    2022

  36. [44]

    , Oh , Sang-Yun S.-Y

    barticle [author] Zapata , Javier J. , Oh , Sang-Yun S.-Y. Petersen , Alexander A. ( 2022 ). Partial separability and functional graphical models for multivariate Gaussian processes . Biometrika 109 665--681 . barticle

    2022

  37. [45]

    Li , Yehua Y

    barticle [author] Zhang , Haozhe H. Li , Yehua Y. ( 2022 ). Unified principal component analysis for sparse and dense functional data under spatial dependency . Journal of Business & Economic Statistics 40 1523--1537 . barticle

    2022

  38. [46]

    , Zhu , Zhengyuan Z

    barticle [author] Zhang , Haozhe H. , Zhu , Zhengyuan Z. Yin , Shuiqing S. ( 2019 ). Identifying precipitation regimes in China using model-based clustering of spatial functional data . arXiv preprint arXiv:1904.10152 . barticle

    2019 arXiv

  39. [47]

    , Shao , Xiaofeng X

    barticle [author] Zhang , Xianyang X. , Shao , Xiaofeng X. , Hayhoe , Katharine K. Wuebbles , Donald J D. J. ( 2011 ). Testing the structural stability of temporally dependent functional observations and application to climate projections . Electronic Journal of Statistics 5 1...

    2011

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.