Pith. sign in

REVIEW 1 major objections 4 minor 28 references

Circular Expectiles

T0 review · 1 major / 4 minor · reviewed 2026-07-10 · grok-4.5

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

Pith's one-line read Circular expectiles extend asymmetric quadratic location to directional data, with uniqueness, consistency, and joint asymptotic normality for positive-density distributions on the circle.

desk verdict Solid, self-contained introduction of circular expectiles with careful non-convex uniqueness proofs and a usable CLT; genuine new location functional for directional data, not a paradigm shift. read the letter →

arxiv 2607.08306 v1 pith:H7A266FU submitted 2026-07-09 stat.ME

Bernhard Klar This is my paper
classification stat.ME MSC 62H1162G0562E20
keywords circularstatisticsdirectionaldatameanexpectilesasymmetriclossidentificationfunctionsasymptoticnormality
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 defines circular expectiles as location functionals for angles by minimizing an asymmetric chord-distance loss. The construction recovers the classical circular mean when the asymmetry parameter is one-half, and otherwise tilts the criterion so that deviations on one side of a reference direction are weighted more heavily than those on the other. Because the circular loss is non-convex, existence and uniqueness are not automatic; the author proves that every distribution with positive density almost everywhere on the circle admits a unique circular expectile. The sample version, linearized about the sample circular mean, is likewise unique almost surely, strongly consistent, and jointly asymptotically normal at any finite collection of asymmetry levels. These properties open a direct route to expectile-based measures of circular dispersion, skewness, and symmetry.

What carries the argument

The circular identification function I_α(ν,F), obtained as the derivative of the asymmetric chord-loss objective, whose unique sign-changing zero is the circular expectile; the same object supplies the estimating equation for the empirical version and the influence functions in the central-limit theorem.

What would settle it

Construct a continuous circular density that vanishes on a positive-measure arc yet still has positive mean resultant length, and check whether the asymmetric chord-loss objective then possesses more than one local minimizer for some asymmetry level.

Watch

Extended reading notes

Core claim

For absolutely continuous circular distributions with positive density almost everywhere and positive mean resultant length, the asymmetric chord-loss criterion admits a unique minimizer for every asymmetry level; the empirical version based on the sample circular mean is unique almost surely, strongly consistent, and jointly asymptotically normal.

Load-bearing premise

The uniqueness and asymptotic proofs require the density to be positive almost everywhere on the whole circle (and continuous at the antipode), so that the identification function has exactly one sign change.

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

1 major / 4 minor

Summary. The paper introduces circular expectiles as minimizers of an asymmetric chord-distance loss on the circle, reducing to the circular mean at α=1/2. Because the criterion is non-convex, the authors develop identification-function representations, a geometric collinearity characterization, and a detailed case analysis proving existence and uniqueness for distributions with positive density a.e. and positive mean resultant length (Theorem 3 and Remark 4). The empirical version is defined by linearizing the circle at the sample circular mean; uniqueness a.s., strong consistency, and finite-dimensional asymptotic normality via a Z-estimator expansion that accounts for the random branch cut are established (Theorems 8 and 13, Proposition 12). Potential applications to interexpectile ranges, skewness profiles, and symmetry diagnostics are sketched, with Monte-Carlo checks and public code.

Significance. The work fills a clear gap between the well-developed theory of linear expectiles and the existing circular location literature (means, medians, quantiles, depths). The non-convexity of the circular criterion makes the uniqueness and asymptotic arguments non-routine; the careful sign-change analysis of the identification function and the explicit treatment of the random branch cut in the CLT are technically solid contributions. Strengths include fully written proofs, an explicit O(n) computational algorithm, a GitHub implementation, and confirmatory Monte-Carlo tables/figures that match the limiting covariance. If the results hold under the stated assumptions, circular expectiles become a usable family of location functionals for directional data and open natural routes to dispersion, skewness and symmetry diagnostics.

major comments (1)
  1. The uniqueness and CLT statements (Theorem 3, Theorem 13) are proved under the strong assumption that the density is positive a.e. on the whole circle (and continuous at the antipode). Example 7 already shows that uniqueness can hold without full support, and Remark 15 notes that the CLT needs only uniqueness of roots, positive slopes A_j, consistency and antipodal continuity. The main theorems should be restated under the weaker conditions actually used in the proofs, or a short corollary should record the minimal hypotheses; otherwise the stated scope is unnecessarily restrictive relative to the arguments given.
minor comments (4)
  1. In Definition 1 and the subsequent empirical construction the dependence on a reference direction (population or sample circular mean) is essential for the induced linear order; a brief remark comparing this choice with the reference used for projection quantiles (Ley et al., 2014) would help readers place the construction.
  2. Figure 1 and Figure 2 would benefit from explicit legends or captions that list the α-levels and sample sizes more clearly; the current stack-plot layout is informative but dense.
  3. The applications subsection (4.1) is suggestive rather than developed; even a short numerical illustration of the interexpectile range or the skewness profile γ_α on a standard circular distribution would strengthen the closing claims.
  4. A few typographical slips remain (e.g., spacing around α=1/2, occasional line-break artefacts in displayed equations). A careful copy-edit pass would remove them.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; uniqueness, consistency and CLT are derived from first-order analysis of the chord-loss identification function under stated density assumptions.

full rationale

The paper defines circular expectiles via an asymmetric chord-distance criterion that reduces to the classical circular mean at α=1/2, then proves existence/uniqueness by direct sign-change analysis of the identification function I_α (Theorem 3, with case splits on C_α(-π/2)), uniqueness of the empirical version by the same analysis on ordered samples (Theorem 8), strong consistency by a sign-change argument after replacing the branch center by the sample mean (Proposition 12), and finite-dimensional asymptotic normality via a standard Z-estimator expansion that accounts for the branch-cut derivative (Theorem 13). All steps are self-contained once the classical circular mean and linear-expectile background are granted; self-citations (Holzmann-Klar 2016, Eberl-Klar series) supply only linear-case analogies and are not load-bearing for the circular claims. No free parameters are fitted to data and then re-presented as predictions; Monte-Carlo tables and figures are purely confirmatory. The positive-density assumption is stronger than necessary (as the paper itself notes in Example 7 and Remark 15) but does not create circularity.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

The central claims rest on standard measure-theoretic probability, classical properties of the circular mean, and the definition of linear expectiles; the only paper-specific modeling choices are the chord-based asymmetric loss and the use of the sample circular mean as branch center. No numerical free parameters are fitted; the invented entity is the circular expectile itself, which is given an independent computational and asymptotic handle.

assumptions (4)
  • domain assumption The circular mean is the unique minimizer of E[1-cos(X-ν)] whenever the mean resultant length R>0.
    Used throughout as the reference direction that linearizes the circle (Section 1.1 and Definition 1).
  • domain assumption Linear expectiles are the unique minimizers of the asymmetric quadratic loss and are identified by the first-order condition αE(X-e)+ = (1-α)E(X-e)-.
    Background recalled in Section 1.2; the circular construction is the direct analogue.
  • standard math Standard Z-estimator theory (van der Vaart, Thm 5.21) applies once the estimating function is Donsker and L2-continuous.
    Invoked for the finite-dimensional CLT in the proof of Theorem 13.
  • ad hoc to paper The density is positive a.e. on (-π,π) and continuous at the antipode.
    Imposed for uniqueness (Thm 3) and for the branch-cut derivative in the CLT (Thm 13); stronger than necessary as Example 7 shows.
invented entities (1)
  • circular α-expectile μ_α independent evidence
    purpose: Asymmetric location functional on the circle that interpolates the circular mean and yields dispersion/skewness diagnostics.
    Defined as the unique minimizer of the asymmetric chord loss; independent computational algorithm and asymptotic theory are supplied, so the entity is not free-floating.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Circular Expectiles." pith.science (2026). https://pith.science/paper/H7A266FU

@misc{pith2026260708306,
  author       = {Pith},
  title        = {Pith review of: Circular Expectiles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H7A266FU}},
  note         = {Machine review of arXiv:2607.08306}
}
abstract

In this work, we introduce circular expectiles as minimizers of an asymmetric circular loss function based on chord distance. In contrast to the linear expectile criterion, the resulting circular optimization problem is non-convex, so existence and uniqueness require a separate analysis. The construction extends linear expectiles to directional data while preserving the circular mean as the symmetric case corresponding to $\alpha=1/2$. We derive basic representations of the objective function and the associated identification function, and give a geometric interpretation that generalizes the corresponding representation for the circular mean. Furthermore, we prove the existence and uniqueness of the minimizers for distributions with positive density on the circle. The empirical circular expectile is defined by using the sample circular mean as reference direction for the induced linear order on the circle. We prove the uniqueness of the empirical expectile, as well as its consistency and finite-dimensional asymptotic normality. Finally, we indicate possible applications to circular measures of dispersion, skewness, and symmetry diagnostics.

Figures

Figures reproduced from arXiv: 2607.08306 by the authors.

Figure 1
Figure 1. Empirical lower and upper circular α-expectiles for α = 0.5 (red), α = 0.25 (green), 0.1 (orange), 0.05 (yellow), 0.01 (brown) for samples of size 100 from the von Mises distribution vM(0, κ) with κ = 3 (left) and κ = 1 (right). replacing the true branch center µ by ˆµn does not affect the empirical identification function asymptotically. Proposition 12 (Consistency of the empirical circular expectile). Let X1, X2, … view at source ↗
Figure 2
Figure 2. Stack plot of ˆµn (left), ˆµn,0.75 (middle) and ˆµn,0.90 (right) for samples of sizes 25 (upper row), 100 (middle row) and 400 (lower row) from the von Mises distribution vM(0, 1), with 104 replications. 4.1. Potential applications The results developed above suggest several possible applications of circular expectiles. Throughout this subsection, angular differences are understood in the µ-centered chart (µ − π, µ … view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    and Lund, U

    Agostinelli, C. and Lund, U. (2025). R package circular : Circular Statistics (version 0.5-2)

    work page 2025

  2. [2]

    Bellini, F., Fadina, T., Wang, R., and Wei, Y. (2022). Parametric measures of variability induced by risk measures. Insurance: Mathematics and Economics , 106:270--284

    work page 2022

  3. [3]

    Bellini, F., Klar, B., and M \"u ller, A. (2018). Expectiles, omega ratios and stochastic ordering. Methodology and Computing in Applied Probability , 20:855--873

    work page 2018

  4. [4]

    Bellini, F., Klar, B., M \"u ller, A., and Gianin, E. R. (2014). Generalized quantiles as risk measures. Insurance: Mathematics and Economics , 54:41--48

    work page 2014

  5. [5]

    and Patrangenaru, V

    Bhattacharya, R. and Patrangenaru, V. (2003). Large sample theory of intrinsic and extrinsic sample means on manifolds . The Annals of Statistics , 31(1):1 -- 29

    work page 2003

  6. [6]

    and Patrangenaru, V

    Bhattacharya, R. and Patrangenaru, V. (2005). Large sample theory of intrinsic and extrinsic sample means on manifolds—II . The Annals of Statistics , 33(3):1225 -- 1259

    work page 2005

  7. [7]

    Di Marzio , M., Panzera, A., and Taylor, C. C. (2016). Nonparametric circular quantile regression. Journal of Statistical Planning and Inference , 170:1--14

    work page 2016

  8. [8]

    and Klar, B

    Eberl, A. and Klar, B. (2020). Asymptotic distributions and performance of empirical skewness measures. Computational Statistics & Data Analysis , 146:106939

    work page 2020

Show all 28 references
  1. [9]

    and Klar, B

    Eberl, A. and Klar, B. (2022). Expectile-based measures of skewness. Scandinavian Journal of Statistics , 49(1):373--399

    2022

  2. [10]

    and Klar, B

    Eberl, A. and Klar, B. (2023). Stochastic orders and measures of skewness and dispersion based on expectiles. Statistical Papers , 64:509--527

    2023

  3. [11]

    Fisher, N. I. (1985). Spherical medians. Journal of the Royal Statistical Society. Series B (Methodological) , 47(2):342--348

    1985

  4. [12]

    Fisher, N. I. (1993). Statistical Analysis of Circular Data . Cambridge University Press

    1993

  5. [13]

    and Klar, B

    Holzmann, H. and Klar, B. (2016). Expectile asymptotics. Electronic Journal of Statistics , 10(2):2355--2371

    2016

  6. [14]

    Hotz, T. (2013). Extrinsic vs intrinsic means on the circle. In Nielsen, F. and Barbaresco, F., editors, Geometric Science of Information , pages 433--440. Springer Berlin Heidelberg

    2013

  7. [15]

    and Huckemann, S

    Hotz, T. and Huckemann, S. (2015). Intrinsic means on the circle: uniqueness, locus and asymptotics. Annals of the Institute of Statistical Mathematics , 67:177–193

    2015

  8. [16]

    Jammalamadaka, S. R. and SenGupta, A. (2001). Topics in Circular Statistics . World Scientific

    2001

  9. [17]

    a tschmer, V. and Z \

    Kr \"a tschmer, V. and Z \"a hle, H. (2017). Statistical inference for expectile-based risk measures. Scandinavian Journal of Statistics , 44(2):425--454

    2017

  10. [18]

    Ley, C., Sabbah, C., and Verdebout, T. (2014). A new concept of quantiles for directional data and the angular mahalanobis depth. Electronic Journal of Statistics , 8(1):795--816

    2014

  11. [19]

    Liu, R. Y. and Singh, K. (1992). Ordering directional data: Concepts of data depth on circles and spheres. The Annals of Statistics , 20(3):1468--1484

    1992

  12. [20]

    and Jupp, P

    Mardia, K. and Jupp, P. (2000). Directional Statistics. John Wiley & Sons, Ltd, United States

    2000

  13. [21]

    G., Quinn, B

    McKilliam, R. G., Quinn, B. G., and Clarkson, I. V. L. (2012). Direction estimation by minimum squared arc length. IEEE Transactions on Signal Processing , 60(5):2115--2124

    2012

  14. [22]

    Newey, W. K. and Powell, J. L. (1987). Asymmetric least squares estimation and testing. Econometrica: Journal of the Econometric Society , pages 819--847

    1987

  15. [23]

    Pewsey, A. (2004). The large-sample joint distribution of key circular statistics. Metrika , 60:25–32

    2004

  16. [24]

    Purkayastha, S. (1995). An almost sure representation of sample circular median. Journal of Statistical Planning and Inference , 46(1):77--91

    1995

  17. [25]

    R: A Language and Environment for Statistical Computing

    R Core Team (2025). R: A Language and Environment for Statistical Computing . R Foundation for Statistical Computing, Vienna, Austria

    2025

  18. [26]

    and Kneib, T

    Sobotka, F. and Kneib, T. (2012). Geoadditive expectile regression. Computational Statistics & Data Analysis , 56(4):755--767

    2012

  19. [27]

    Van der Vaart, A. (1998). Asymptotic Statistics. Cambridge University Press

    1998

  20. [28]

    Ziegel, J. F. (2016). Coherence and elicitability. Mathematical Finance , 26(4):901--918

    2016

Pith tools

Reviewed July 10, 2026 · model on record in the stance chip above.