Pith. sign in

REVIEW 3 minor 72 references

Polar Depth for Potentially Heavy-Tailed Data

T0 review · 0 major / 3 minor · reviewed 2026-06-28 · grok-4.3

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

Pith's one-line read Polar depth of observations beyond growing thresholds converges to the depth of the limiting distribution for regularly varying vectors.

desk verdict Polar depth is a new notion tailored to regularly varying extremes that converges to the limit depth, with the main value in its fit to polar coordinates for anomaly detection among tails. read the letter →

arxiv 2606.00343 v1 pith:IQ33WAVS submitted 2026-05-29 math.ST stat.COstat.MLstat.TH

Stephan Clemençon , Carlos Fernándes , Pavlo Mozharovskyi , Anne Sabourin This is my paper · ORCID
classification math.STstat.COstat.MLstat.TH
keywords polardepthheavy-taileddistributionsmultivariateextremesstatisticalanomalydetectionregularvariationconvergencetailordering
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

The paper introduces polar depth as a new statistical depth function defined in polar coordinates to match the structure of limiting distributions for multivariate heavy-tailed random variables after marginal normalization. It establishes that the polar depth values computed on the largest observations, those whose Euclidean norm exceeds a threshold t, approach the polar depth of the limiting measure as t tends to infinity. This convergence supplies a threshold-independent way to rank and compare extreme points. The construction is motivated by the need to analyze tails where standard depths such as halfspace depth lose relevance, especially when the support lies in a half-space. Estimation procedures and finite-sample behavior are examined alongside numerical illustrations focused on anomaly detection within extremes.

What carries the argument

Polar depth function, a depth measure defined in polar coordinates that orders points according to their position relative to the angular measure of the limiting distribution.

What would settle it

Generate repeated samples from a regularly varying multivariate distribution with known limiting angular measure; compute the empirical polar depths of points with norm exceeding successively larger thresholds and verify whether these depths approach the known limiting polar depth values within sampling error.

Watch

Extended reading notes

Core claim

The polar depth function, expressed directly in polar coordinates to align with the angular and radial components of the limiting distribution of a regularly varying random vector after appropriate marginal normalization, satisfies the property that the polar depth of all observations with norm larger than t converges to the polar depth of that limiting distribution as t tends to infinity.

Load-bearing premise

The underlying random vector is regularly varying with marginals that admit suitable normalization so that a non-degenerate limiting distribution exists.

Editorial extensions

If this is right

  • Extreme observations can be ordered by depth in a manner that becomes stable for large thresholds.
  • Anomaly detection among tail points becomes consistent across choices of threshold.
  • Polar depth remains meaningful for distributions supported on a half-space where halfspace depth is less informative.
  • Statistical estimation of polar depth from finite samples is feasible both in finite-sample and asymptotic regimes.
  • Numerical experiments confirm the ordering and detection properties for heavy-tailed data.

Reading between the lines

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

  • The same convergence argument could be examined for other depth notions adapted to polar representations of extremes.
  • In practice one could test the rate of convergence by plotting polar depth values against increasing thresholds on real heavy-tailed data sets such as financial returns.
  • The method suggests a natural way to define depth-based confidence regions that grow with the tail threshold.
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

0 major / 3 minor

Summary. The paper introduces Polar Depth, a statistical depth notion for multivariate heavy-tailed distributions expressed in polar coordinates to align with the limiting distribution of regularly varying random vectors after marginal normalization. The central result is a convergence theorem: under appropriate assumptions, the polar depth of observations with norm exceeding t converges to the polar depth of the limiting distribution as t → ∞. The manuscript establishes properties of this depth, analyzes its estimation from finite-sample and asymptotic viewpoints, discusses applications to anomaly detection among extremes, and includes numerical results to illustrate relevance.

Significance. If the convergence result holds, the work supplies a depth measure specifically adapted to heavy-tailed extremes and halfspace-supported distributions, where alternatives such as halfspace depth are less suitable. The explicit convergence to the limiting polar depth provides theoretical grounding for applying the notion to large observations, while the finite-sample and asymptotic estimation analysis plus numerical demonstrations add practical value for extreme-value analysis and outlier detection. The combination of a proved convergence statement with empirical validation is a clear strength.

minor comments (3)
  1. [Abstract] Abstract: the claim that polar depth 'finds natural applications in anomaly detection' would benefit from a one-sentence pointer to the specific numerical experiment or figure that demonstrates this use.
  2. [Convergence theorem (likely §3)] The normalization of marginals invoked for the limiting distribution should be stated with an explicit reference to the relevant assumption or equation in the convergence theorem section, even if the proof is otherwise self-contained.
  3. [Numerical results] Numerical results section: report the precise sample sizes, dimensions, and number of Monte Carlo replications used, to allow readers to gauge the reliability of the finite-sample illustrations.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and accurate summary of our manuscript introducing Polar Depth for heavy-tailed data. The recommendation of minor revision is noted, and we appreciate the recognition of the convergence result, estimation analysis, and applications to anomaly detection. No major comments were raised in the report.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity detected

full rationale

The paper defines polar depth in polar coordinates aligned with the limiting distribution of regularly varying vectors (after marginal normalization) and proves convergence of the depth for ||X|| > t as t → ∞. This is a standard limit theorem under explicit assumptions; the derivation does not reduce any claimed prediction or uniqueness result to a fitted parameter, self-citation chain, or definitional tautology. No load-bearing steps match the enumerated circularity patterns.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

The central claims rest on the assumption of regular variation for the multivariate distribution and appropriate marginal normalization to obtain the limiting extreme value behavior; no free parameters or invented entities are explicitly introduced beyond the new depth definition itself.

assumptions (1)
  • domain assumption The random variable is regularly varying, with marginals appropriately normalized, so that a limiting distribution exists beyond large thresholds.
    Stated in the motivation and used for the convergence result.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Polar Depth for Potentially Heavy-Tailed Data." pith.science (2026). https://pith.science/paper/IQ33WAVS

@misc{pith2026260600343,
  author       = {Pith},
  title        = {Pith review of: Polar Depth for Potentially Heavy-Tailed Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IQ33WAVS}},
  note         = {Machine review of arXiv:2606.00343}
}
read the original abstract

Motivated by the analysis of the behaviour of extremes from multivariate heavy-tailed distributions, we introduce a novel notion of statistical depth, referred to as Polar Depth. The polar depth function is naturally expressed in polar coordinates, as is the limiting distribution of a regularly varying random variable, beyond asymptotically large thresholds, once its marginals have been appropriately normalized. Not only does the polar depth function make it easy to order the extreme values taken by a heavy-tailed random variable X and finds natural applications in anomaly detection, but it is also possible to show, as we prove it under appropriate assumptions in this article, that the polar depth of the largest observations, i.e. observations X which norm is larger than t>0, converges to the polar depth of the limiting distribution as t converges to infinity. Although designed to quantify the depth of multivariate extremes, the polar depth is interesting in its own right, insofar as this notion is more relevant for distributions whose support is included in a halfspace than the alternatives proposed in the literature, the halfspace depth in particular. Here, we demonstrate its properties and analyze statistical issues related to its estimation from both finite-sample and asymptotic points of view. We present numerical results to empirically demonstrate its relevance, particularly for the statistical analysis of extreme observations and more specifically for the identification of anomalies among them.

Figures

Figures reproduced from arXiv: 2606.00343 by the authors.

Figure 1
Figure 1. Illustrative sample of 10 realizations of the heavy tailed r.v. [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Level sets for points in the bulk of the distribution. [PITH_FULL_IMAGE:figures/full_fig_p024_2.png] view at source ↗
Figure 3
Figure 3. Level set at level 1/n = 0.0002. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Boxplots of the ratio of the estimated and population depth, for different population level sets [PITH_FULL_IMAGE:figures/full_fig_p025_4.png]
Figure 5
Figure 5. Figure 5: ROC curves for the anomaly detection task in the bulk of the distribution, for a sample of [PITH_FULL_IMAGE:figures/full_fig_p026_5.png]
Figure 6
Figure 6. Figure 6: ROC curves for the anomaly detection task in the tail of the distribution, for a sample of [PITH_FULL_IMAGE:figures/full_fig_p027_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 1 canonical work pages

  1. [1]

    Beirlant, Y

    J. Beirlant, Y. Goegebeur, J. Segers, and J. L. Teugels.Statistics of extremes: theory and appli- cations. John Wiley & Sons, 2006

    2006

  2. [2]

    Billingsley and F

    P. Billingsley and F. Topsøe. Uniformity in weak convergence.Zeitschrift f¨ ur Wahrscheinlichkeit- stheorie und verwandte Gebiete, 7(1):1–16, 1967

    1967

  3. [3]

    N. H. Bingham, C. M. Goldie, and J. L. Teugels.Regular variation. Encyclopedia of mathematics and its applications 27. Cambridge University Press, 1987. 27

    1987

  4. [4]

    Boucheron, O

    S. Boucheron, O. Bousquet, and G. Lugosi. Theory of classification: A survey of some recent advances.ESAIM: probability and statistics, 9:323–375, 2005

    2005

  5. [5]

    Boucheron and M

    S. Boucheron and M. Thomas. Tail index estimation, concentration and adaptivity.Electronic Journal of Statistics, pages 2751 – 2792, 2015

    2015

  6. [6]

    V.-E. Brunel. Concentration of the empirical level sets of tukey’s halfspace depth.Probability Theory and Related Fields, 173(3):1165–1196, 2019

    2019

  7. [7]

    Bullen.Dictionary of inequalities

    P. Bullen.Dictionary of inequalities. CRC Press, 1 edition, 1998

    1998

  8. [8]

    M. A. Burr and R. J. Fabrizio. Uniform convergence rates for halfspace depth.Statistics and Probability Letters, 124:33–40, 2017

    2017

Show all 72 references
  1. [9]

    J.-J. Cai, J. H. Einmahl, and L. De Haan. Estimation of extreme risk regions under multivariate regular variation. 2011

    2011

  2. [10]

    Carpentier and A

    A. Carpentier and A. K. H. Kim. Adaptive and minimax optimal estimation of the tail coefficient. Statistica Sinica, 25(3):1133–1144, 2015

    2015

  3. [11]

    Chaudhuri

    P. Chaudhuri. On a geometric notion of quantiles for multivariate data.Journal of the American Statistical Association, 91(434):862–872, 1996

    1996

  4. [12]

    Chernozhukov, A

    V. Chernozhukov, A. Galichon, M. Hallin, and M. Henry. Monge–kantorovich depth, quantiles, ranks and signs.The Annals of Statistics, 45(1):223–256, 02 2017

    2017

  5. [13]

    Chiapino, S

    M. Chiapino, S. Cl´ emen¸ con, V. Feuillard, and A. Sabourin. A multivariate extreme value theory approach to anomaly clustering and visualization.Computational Statistics, 35(2):607–628, 2020

    2020

  6. [14]

    Cl´ emen¸ con, N

    S. Cl´ emen¸ con, N. Huet, and A. Sabourin. On regression in extreme regions.Electronic Journal of Statistics, 19(2):4784–4828, 2025

    2025

  7. [15]

    Cl´ emen¸ con, H

    S. Cl´ emen¸ con, H. Jalalzai, S. Lhaut, A. Sabourin, and J. Segers. Concentration bounds for the empirical angular measure with statistical learning applications.Bernoulli, 29(4):2797 – 2827, 2023

    2023

  8. [16]

    Cl´ emen¸ con, P

    S. Cl´ emen¸ con, P. Mozharovskyi, and G. Staerman. Affine invariant integrated rank-weighted statistical depth: properties and finite sample analysis.Electronic Journal of Statistics, 17(2):3854 – 3892, 2023

    2023

  9. [17]

    Coles, J

    S. Coles, J. Bawa, L. Trenner, and P. Dorazio.An introduction to statistical modeling of extreme values, volume 208. Springer, 2001

    2001

  10. [18]

    De Haan and A

    L. De Haan and A. Ferreira.Extreme value theory: an introduction. Springer Science & Business Media, 2007

    2007

  11. [19]

    Devroye, L

    L. Devroye, L. Gy¨ orfi, and G. Lugosi.A probabilistic theory of pattern recognition, volume 31. Springer Science & Business Media, 2013

    2013

  12. [20]

    D. L. Donoho.Breakdown Properties of Multivariate Location Estimators. PhD thesis, Harvard University, 1982

    1982

  13. [21]

    D. L. Donoho and M. Gasko. Breakdown properties of location estimates based on halfspace depth and projected outlyingness.The Annals of Statistics, 20:1803–1827, 1992

    1992

  14. [22]

    Dyckerhoff

    R. Dyckerhoff. Data depths satisfying the projection property.AStA - Advances in Statistical Analysis, 88(2):163–190, 2004. 28

    2004

  15. [23]

    Dyckerhoff, C

    R. Dyckerhoff, C. Ley, and D. Paindaveine. Depth-based runs test for bivariate central symmetry. Annals of the Institute of Statistical Mathematics, 67:917–941, 2015

    2015

  16. [24]

    Dyckerhoff and P

    R. Dyckerhoff and P. Mozharovskyi. Exact computation of the halfspace depth.Computational Statistics & Data Analysis, 98:19–30, 2016

    2016

  17. [25]

    Dyckerhoff, P

    R. Dyckerhoff, P. Mozharovskyi, and S. Nagy. Approximate computation of projection depths. Computational Statistics & Data Analysis, 157(C), 2021

    2021

  18. [26]

    Dyckerhoff and S

    R. Dyckerhoff and S. Nagy. Exact computation of angular halfspace depth.Statistics and Com- puting, 35(6), Aug. 2025

    2025

  19. [27]

    J. H. Einmahl, J. Li, and R. Y. Liu. Bridging centrality and extremity: Refining empirical data depth using extreme value statistics.Annals of Statistics, 43(6):2738–2765, 2015

    2015

  20. [28]

    J. H. J. Einmahl, L. de Haan, and V. I. Piterbarg. Nonparametric estimation of the spectral measure of an extreme value distribution.The Annals of Statistics, 29(5):1401–1423, 2001

    2001

  21. [29]

    J. H. J. Einmahl and J. Segers. Maximum empirical likelihood estimation of the spectral measure of an extreme-value distribution.The Annals of Statistics, 37(5B):2953–2989, 2009

    2009

  22. [30]

    K. W. Fang.Symmetric Multivariate and Related Distributions. Chapman & Hall, 1990

    1990

  23. [31]

    Fernandez and S

    C. Fernandez and S. Cl´ emencon. Anomaly detection based on Markov data: a statistical depth approach. InProceedings of the LOD Conference, 2025

    2025

  24. [32]

    C. A. Fern´ andez, P. Bertail, and S. Cl´ emen¸ con. Tail index estimation for discrete heavy-tailed distributions with application to statistical inference for regular Markov chains.TEST, pages 1–23, 2025

    2025

  25. [33]

    He and J

    Y. He and J. H. Einmahl. Estimation of extreme depth-based quantile regions.Journal of the Royal Statistical Society Series B: Statistical Methodology, 79(2):449–461, 2017

    2017

  26. [34]

    Hult and F

    H. Hult and F. Lindskog. Regular variation for measures on metric spaces.Publications de l’Institut Math´ ematique, 80(94):121–140, 2006

    2006

  27. [35]

    Koshevoy and K

    G. Koshevoy and K. Mosler. Zonoid trimming for multivariate distributions.The Annals of Statistics, 25(5):1998–2017, 10 1997

    1998

  28. [36]

    G. A. Koshevoy. The Tukey depth characterizes the atomic measure.Journal of Multivariate Analysis, 83:360–364, 2002

    2002

  29. [37]

    Lafaye De Micheaux, P

    P. Lafaye De Micheaux, P. Mozharovskyi, and M. Vimond. Depth for curve data and applications. Journal of the American Statistical Association, 116(536):1881–1897, 2022

    2022

  30. [38]

    Lederer, A

    J. Lederer, A. Sabourin, and M. Taheri. Adaptive tail index estimation: minimal assumptions and non-asymptotic guarantees, 2025

    2025

  31. [39]

    Lhaut, A

    S. Lhaut, A. Sabourin, and J. Segers. Uniform concentration bounds for frequencies of rare events. Statistics & Probability Letters, 189:109610, 2022

    2022

  32. [40]

    J. Li, J. A. Cuesta-Albertos, and R. Y. Liu. DD-classifier: Nonparametric classification procedure based on DD-plot.Journal of the American Statistical Association, 107:737–753, 2012

    2012

  33. [41]

    R. Liu, J. Parelius, and K. Singh. Multivariate analysis by data depth: descriptive statistics, graphics and inference,(with discussion and a rejoinder by liu and singh).The Annals of Statistics, 27(3):783–858, 1999. 29

    1999

  34. [42]

    R. Y. Liu. On a notion of data depth based upon random simplices.The Annals of Statistics, 1990

    1990

  35. [43]

    R. Y. Liu.Data Depth and Multivariate Rank Tests, page 279–294. North-Holland, Amsterdam, 1992

    1992

  36. [44]

    R. Y. Liu and K. Singh. A quality index based on data depth and multivariate rank tests.Journal of the American Statistical Association, 88(421):252–260, 1993

    1993

  37. [45]

    X. Liu, K. Mosler, and P. Mozharovskyi. Fast computation of tukey trimmed regions and median in dimensionp >2.Journal of Computational and Graphical Statistics, 2018. in press

    2018

  38. [46]

    Liu and Y

    X. Liu and Y. Zuo. Computing halfspace depth and regression depth.Communications in Statistics - Simulation and Computation, 2014

    2014

  39. [47]

    G. Lugosi. Pattern classification and learning theory. InPrinciples of nonparametric learning, pages 1–56. Springer, 2002

    2002

  40. [48]

    K. Mosler. Depth statistics. In C. Becker, R. Fried, and S. Kuhnt, editors,Robustness and Complex Data Structures: Festschrift in Honour of Ursula Gather, pages 17–34. Springer, 2013

    2013

  41. [49]

    Mosler and P

    K. Mosler and P. Mozharovskyi. Choosing among notions of multivariate depth statistics.Statis- tical Science, 37(3):348–368, 2022

    2022

  42. [50]

    Mozharovskyi, J

    P. Mozharovskyi, J. Josse, and F. Husson. Nonparametric imputation by data depth.Journal of the American Statistical Association, 115(529):241–523, 2020

    2020

  43. [51]

    Mozharovskyi, K

    P. Mozharovskyi, K. Mosler, and T. Lange. Classifying real-world data with theDDα-procedure. Advances in Data Analysis and Classification, 9:287–314, 2015

    2015

  44. [52]

    Mozharovskyi and R

    P. Mozharovskyi and R. Valla. Anomaly detection using data depth: multivariate case.Interna- tional Journal of Data Science and Analytics, 116:5171–5196, 2025

    2025

  45. [53]

    S. Nagy, H. Demni, D. Buttarazzi, and G. C. Porzio. Theory of angular depth for classification of directional data.Advances in Data Analysis and Classification, 2023

    2023

  46. [54]

    S. Nagy, R. Dyckerhoff, and P. Mozharovskyi. Uniform convergence rates for the approximated halfspace and projection depth.Electronic Journal of Statistics, 14(2):3939 – 3975, 2020

    2020

  47. [55]

    S. Nagy, I. Gijbels, M. Omelka, and D. Hlubinka. Integrated depth for functional data: statistical properties and consistency.ESAIM: PS, 20:95–130, 2016

    2016

  48. [56]

    Nagy and P

    S. Nagy and P. Laketa. Theoretical properties of angular halfspace depth.Bernoulli, 31(2):1007– 1031, 2025

    2025

  49. [57]

    S. Nagy, P. Laketa, and R. Dyckerhoff. Angular halfspace depth: computation. In G. C. Porzio, C. Rampichini, and C. Bocci, editors,CLADAG 2021. Book of Abstracts and Short Papers, pages 169–172. Firenze University Press, 2021

    2021

  50. [58]

    H. Oja. Descriptive statistics for multivariate distributions.Statistics & Probability Letters, 1(6):327 – 332, 1983

    1983

  51. [59]

    Reiss.Approximate distributions of order statistics: with applications to nonparametric statistics

    R.-D. Reiss.Approximate distributions of order statistics: with applications to nonparametric statistics. Springer science & business media, 2012

    2012

  52. [60]

    S. I. Resnick.Heavy-tail phenomena: probabilistic and statistical modeling. Springer Science & Business Media, 2007. 30

    2007

  53. [61]

    S. I. Resnick.Extreme values, regular variation, and point processes, volume 4. Springer Science & Business Media, 2008

    2008

  54. [62]

    P. J. Rousseeuw and A. Struyf. Computing location depth and regression depth in higher dimen- sions.Statistics and Computing, 8(3):193–203, 1998

    1998

  55. [63]

    S. N. Samuel Kotz.Multivariate T-Distributions and Their Applications. Cambridge University Press, 2004

    2004

  56. [64]

    Staerman, E

    G. Staerman, E. Adjakossa, P. Mozharovskyi, V. Hofer, J. S. Gupta, and S. Cl´ emen¸ con. Functional anomaly detection: a benchmark study.International Journal of Data Science and Analytics, 16(16):101–117, 2023

    2023

  57. [65]

    Staerman, P

    G. Staerman, P. Mozharovskyi, and S. Cl´ emencon. The area of the convex hull of sampled curves: a robust functional statistical depth measure. InInternational Conference on Artificial Intelligence and Statistics, pages 570–579. PMLR, 2020

    2020

  58. [66]

    Struyf and P

    A. Struyf and P. Rousseeuw. Halfspace depth and regression depth characterize the empirical distribution.Journal of Multivariate Analysis, 69:135–153, 1999

    1999

  59. [67]

    J. W. Tukey. Mathematics and the picturing of data. In R. James, editor,Proceedings of the International Congress of Mathematicians, volume 2, pages 523–531, 1975

    1975

  60. [68]

    Valla, P

    R. Valla, P. Mozharovskyi, and F. d’Alch´ e Bud. Abnormal component analysis.Technometrics, In press, 2025

    2025

  61. [69]

    Vardi and C.-H

    Y. Vardi and C.-H. Zhang. The multivariate l1-median and associated data depth.Proceedings of the National Academy of Sciences, 97(4):1423–1426, 2000

    2000

  62. [70]

    J. Virta. Spatial depth for data in metric spaces.arXiv preprint arXiv:2306.09740, 2023

    2023

  63. [71]

    Zuo and R

    Y. Zuo and R. Serfling. General notions of statistical depth function.The Annals of Statistics, 28(2):461–482, 2000

    2000

  64. [72]

    Zuo and R

    Y. Zuo and R. Serfling. Structural properties and convergence results for contours of sample statistical depth functions.The Annals of Statistics, 28(2):483–499, 2000. 31 A Proofs and Intermediate Results A.1 Proofs of Results from Section 3 A.1.1 Proof of Lemma 3.1 From the d...

    2000

Pith tools

Reviewed June 28, 2026 · model on record in the stance chip above.