Pith. sign in

REVIEW 3 major objections 4 minor 26 references

Autocorrelation is usually more sensitive than angular power spectra for spotting medium-scale cosmic-ray anisotropy, and both methods flag a ~10° iron excess above 20 PeV in a KASCADE sample.

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 ↗

T0 review · grok-4.5

2026-07-14 05:48 UTC pith:O2XTH6CZ

load-bearing objection Solid methods ranking of autocorrelation vs power spectrum under partial-sky exposure; the Fe ~10° indication on 10% KASCADE is real but overstated by incomplete trials factor. the 3 major comments →

arxiv 2607.11405 v1 pith:O2XTH6CZ submitted 2026-07-13 astro-ph.HE

Cosmic-ray anisotropy: sensitivity of methods and implications for KASCADE data

M. Yu. Kuznetsov , N. A. Petrov , E. M. Shinkevich This is my paper
classification astro-ph.HE
keywords cosmic-ray anisotropyangular power spectrumautocorrelation functionKASCADEmass compositionmedium-scale structureLandy-Szalay estimator
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.

Ground-based cosmic-ray arrays see only part of the sky with uneven exposure, so the usual dipole-in-right-ascension test gives little information about the true pattern of arrival directions. This paper asks which of two standard tools—the angular power spectrum and the cumulative Landy–Szalay autocorrelation function—is better at recovering medium- and small-scale structure under those observational constraints. Controlled simulations of Gaussian multipole, turbulent-diffusion and single-source maps show that the autocorrelation function yields lower post-trial p-values in most cases; the power spectrum becomes competitive only when a single multipole is strong. Applying both estimators to a 10 % public KASCADE sample (events classified by a neural network into proton, helium, carbon, silicon and iron groups) produces a consistent >2.5σ excess at roughly 10° angular scale for iron nuclei above about 20 PeV. No comparable signal appears for other species or for the all-particle set, and the East–West dipole search remains consistent with isotropy. The result is presented as an unblinded indication that will be tested on the remaining 90 % of the data.

Core claim

Under realistic limited-field-of-view exposure the cumulative Landy–Szalay autocorrelation function is the more sensitive detector of medium- and small-scale cosmic-ray anisotropy for the majority of the models examined, while both that function and the angular power spectrum return a post-trial significance greater than 2.5σ at ~10° for the iron mass group at E ≳ 20 PeV in the 10 % KASCADE sample.

What carries the argument

The cumulative Landy–Szalay estimator w(ψ) = (DD − 2DR + RR)/RR, evaluated on a geometric or time-scrambled exposure map and compared with the multipole-corrected angular power spectrum C_l (and its cumulative version), together with a post-trial p-value that scans only angular scale or multipole.

Load-bearing premise

Energy bins and mass groups are treated as statistically independent, so the look-elsewhere penalty is applied only over angular scale or multipole; correlated reconstruction or exposure systematics would inflate the quoted significance.

What would settle it

Analysis of the remaining 90 % KASCADE events with the same mass classification, exposure reconstruction and post-trial procedure must either reproduce a post-trial excess ≳2.5σ at ~10° for iron above 20 PeV or fail to do so.

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

If this is right

  • Autocorrelation should be preferred as the first-line search statistic for medium-scale cosmic-ray anisotropy in partial-sky experiments.
  • A confirmed iron-specific excess at ~10° would point to rigidity-dependent redistribution of anisotropy by local turbulent magnetic fields.
  • Mass-resolved sky maps become a practical tool once neural-network composition classification is available.
  • The remaining 90 % of KASCADE data can immediately confirm or refute the reported iron signal without new hardware.

Where Pith is reading between the lines

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

  • If the iron excess survives the full sample, similar mass-group analyses at other northern arrays should show the same angular scale near the same rigidity.
  • The failure of the cumulative power spectrum relative to both binned C_l and autocorrelation suggests that simply stacking multipoles does not automatically improve sensitivity under partial-sky mixing.
  • Time-scrambling exposure maps may systematically under-estimate significance for structures larger than a few tens of degrees; a joint analysis with geometric exposure could quantify the bias.

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

3 major / 4 minor

Summary. The paper compares the sensitivity of the cumulative autocorrelation function (Landy–Szalay estimator) and the angular power spectrum (PolSpice-corrected) to medium- and small-scale cosmic-ray anisotropy under limited field-of-view and non-uniform exposure. Using three families of Monte-Carlo flux models (Gaussian multipole, turbulent diffusion, single Gaussian source) with geometric exposure, it finds that autocorrelation is systematically more sensitive for most tested cases, while the power spectrum can add information when the signal is strong and concentrated in multipoles. As an application, both estimators are applied to a 10 % “unblind” sample of KASCADE events (all-particle and five CNN-classified mass groups) after time-scrambling exposure reconstruction; an indication of anisotropy at post-trial p ≈ 0.005–0.009 (∼10° scale) is reported for iron nuclei above ∼20 PeV. Large-scale East–West dipole searches yield no significant signal.

Significance. If the relative-sensitivity ranking holds under realistic exposures, the work supplies a practical, well-documented recommendation for future anisotropy analyses with partial-sky arrays. The explicit comparison of Landy–Szalay versus the standard pair-counting estimator, the introduction of a cumulative power-spectrum statistic, and the controlled Monte-Carlo ensembles (10 000 isotropic realizations, multiple map realizations) are concrete methodological contributions. The KASCADE iron hint, even if only an unblinded 10 % sample, would be of interest if confirmed with the full data set and a proper global trials factor, because mass-group anisotropies at PeV rigidities remain poorly constrained.

major comments (3)
  1. [Sec. 5 / preceding Table 2] Section 5 (paragraph preceding Table 2) and the post-trial definitions in Sec. 3 explicitly state that the look-elsewhere correction is applied only over angular scale ψ < 60° or multipole ℓ > 3; the five mass groups + all-particle sample and the seven nested energy thresholds are declared “independent” and receive no trials factor. The headline claim of an indication at >2.5σ (post-trial p = 0.0053 for autocorrelation and 0.0093 for the power spectrum on Fe, lg(E/eV) > 16.25) therefore rests on an incomplete global p-value. With ∼40 partially correlated searches the true post-trial significance can easily fall below 2σ. Either a full multi-dimensional trials factor (or a conservative Bonferroni/Sidak bound that accounts for the nested energy bins) must be evaluated, or the claim must be rephrased as a pre-trial local excess that awaits the remaining 90 % of the data.
  2. [Sec. 4 and Sec. 5.2–5.3] The iron result is obtained on a deliberately unblinded 10 % subsample whose CNN mass labels are taken from earlier publications without a re-assessment of purity or of possible residual correlations between species induced by the network. Because misclassification fractions of tens of percent are typical at these energies, an artificial clustering signal (or dilution of a real one) cannot be excluded a priori. A quantitative estimate of the leakage of events among mass groups, or at least a statement that the full-data analysis will include a purity-smearing systematic, is required before the >2.5σ indication can be regarded as robust.
  3. [Sec. 4] Time-scrambling is correctly noted to overestimate the isotropic background for extended structures (Sec. 4). For the iron excess at ∼10° this bias works against detection, yet no quantitative bound on the possible underestimation of significance is given. A short Monte-Carlo exercise injecting the observed autocorrelation amplitude into scrambled maps would clarify whether the reported local p-values remain meaningful after this known systematic.
minor comments (4)
  1. [Sec. 5.2 / Fig. 13] Repeated typographical error “post-trail” instead of “post-trial” appears in the caption of Fig. 13 and in the text of Sec. 5.2.
  2. [Sec. 5.3] The cumulative power-spectrum definition (Eq. 3.3) is introduced but then discarded for the data analysis; a one-sentence justification in Sec. 5.3 would improve readability.
  3. [Fig. 8] Figure 8 (right) shows a fit residual χ²/ndof = 166 for the zenith distribution; the functional form “sin cos” is not specified and the large χ² is left uncommented.
  4. [Sec. 4–5] The phrase “unblind dataset” / “unblinded dataset” is used inconsistently; standard terminology is “unblinded”.

Circularity Check

0 steps flagged

No load-bearing circularity; method-sensitivity rankings come from independent Monte-Carlo ensembles and the KASCADE p-values are direct data-to-isotropic comparisons that use prior CNN labels only as input.

full rationale

The paper’s central methodological claim (autocorrelation more sensitive than angular power spectrum for most tested models) is obtained by generating independent isotropic and anisotropic event sets under geometric exposure, injecting known flux models (Gaussian Cl, turbulent cascade Cl ~ 1/[(2l+1)(l+1)(l+2)], single Gaussian source), and ranking post-trial p-values; nothing is fitted to data and then re-predicted. The KASCADE application is presented as an illustration (“as a test of our findings”) that uses the authors’ earlier CNN mass classification only to define event subsets; the subsequent East-West, Landy-Szalay autocorrelation and PolSpice Cl analyses compare those subsets against time-scrambled isotropic backgrounds and therefore do not force the reported p-values by construction. Self-citations to Refs. [22,23] supply the mass labels but are not invoked as uniqueness theorems or as the sole justification of any anisotropy result. No self-definitional loops, fitted-input-as-prediction steps, or ansatz-smuggling appear in the derivation chain. The incomplete look-elsewhere treatment of energy/mass bins is a separate statistical concern, not circularity. Score 1 reflects only the minor, non-load-bearing self-citation for data preparation.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claims rest on standard statistical estimators, geometric exposure models, and the authors’ previously published CNN mass classifier; free parameters are mainly simulation choices (anisotropy amplitude ϵ, multipole binning, Nside) tuned for detectability rather than fitted to the final data claim.

free parameters (4)
  • anisotropy amplitude ϵ = 0.2–0.35 depending on model
    Chosen by hand (0.2–0.35) so that post-trial p-values fall in the 10^{-2}–10^{-3} range for the chosen event statistics; not fitted to real data but controls the sensitivity comparison.
  • multipole bin width Δl=4 and l_max=180 = Δl=4
    Ad-hoc binning used to define both binned and cumulative Cl; affects the cumulative-power-spectrum statistic.
  • Nside=128 HEALPix resolution = 128
    Pixel size ~0.5° chosen smaller than expected anisotropy scales; fixed by hand.
  • zenith-angle cut θ<30° = 30°
    Defines the effective field of view and the post-trial angular range ψ<60°; conventional but free.
axioms (4)
  • domain assumption Isotropic Monte-Carlo sets generated from the geometric or time-scrambled exposure correctly represent the null distribution of both estimators.
    Used throughout Sec. 3 and Sec. 5 to convert observed amplitudes into pre- and post-trial p-values.
  • domain assumption The convolutional neural network mass classification (p, He, C, Si, Fe) from the authors’ prior work is sufficiently pure that residual contamination does not induce spurious medium-scale anisotropy.
    Mass-group results in Sec. 5 rest entirely on this external classifier.
  • domain assumption Time-scrambling preserves the true exposure while erasing only non-physical anisotropies; any residual physical structure is not over-subtracted.
    Sec. 4 explicitly notes the risk of over-estimating background but proceeds with the method.
  • ad hoc to paper Energy bins and primary mass groups may be treated as statistically independent for the purpose of the look-elsewhere correction.
    Stated just before Table 2; reduces the post-trial penalty relative to a full multi-bin scan.

pith-pipeline@v1.1.0-grok45 · 19380 in / 2949 out tokens · 33907 ms · 2026-07-14T05:48:17.008620+00:00 · methodology

0 comments
read the original abstract

We study the problem of measuring the anisotropy of high-energy cosmic rays across all angular scales. The limited field of view and non-uniform exposure of ground-based cosmic-ray experiments reduce their sensitivity to real anisotropy of cosmic-ray arrival directions. A widely used signature of anisotropy -- a dipole of the flux expansion over the right ascension -- provides very limited understanding of the underlying physics of cosmic-ray origin and propagation. In this study we test two other methods: the angular power spectrum and the autocorrelation function for sensitivity to possible medium- and small-scale anisotropies of cosmic-ray flux. We find that the autocorrelation function is the most sensitive estimator for an underlying physical anisotropy in most of the models tested, while the angular power spectrum can provide additional knowledge about cosmic-ray flux properties, when the anisotropy is strong enough. As a test of our findings, we apply these methods to $10\%$ sample of the KASCADE experiment public data. Namely, we consider all-particle set of events and sets of individual mass groups classified by a convolutional neural network. We find an indication of anisotropy at $> 2.5 \sigma$ level at $\sim 10^\circ$ angular scale for the iron nuclei mass group at $E \gtrsim 20$ PeV with both the angular power spectrum and the autocorrelation methods.

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

26 extracted references · 1 canonical work pages

  1. [1]

    Gabici, C

    S. Gabici, C. Evoli, D. Gaggero, P. Lipari, P. Mertsch, E. Orlando et al.,The origin of galactic cosmic rays: Challenges to the standard paradigm,International Journal of Modern Physics D 28(2019) 1930022

    2019

  2. [2]

    Kachelriess and D.V

    M. Kachelriess and D.V. Semikoz,Cosmic Ray Models,Prog. Part. Nucl. Phys.109(2019) 103710 [1904.08160]

    Pith/arXiv arXiv 2019

  3. [3]

    Ginzburg and S.I

    V.L. Ginzburg and S.I. Syrovatsky,Origin of Cosmic Rays,Progress of Theoretical Physics Supplement20(1961) 1. [4]LHAASOcollaboration,Ultrahigh-energy photons up to 1.4 petaelectronvolts from 12γ-ray Galactic sources,Nature594(2021) 33

    1961

  4. [4]

    Ahlers and P

    M. Ahlers and P. Mertsch,Origin of Small-Scale Anisotropies in Galactic Cosmic Rays,Prog. Part. Nucl. Phys.94(2017) 184 [1612.01873]. [6]EAS-TOPcollaboration,Evolution of the cosmic ray anisotropy above 10ˆ14 eV,Astrophys. J. Lett.692(2009) L130 [0901.2740]. [7]Tibet AS-gammacollaboration,Northern sky Galactic Cosmic Ray anisotropy between 10-1000 TeV wit...

    Pith/arXiv arXiv 2017

  5. [5]

    Apel et al.,Search for Large-scale Anisotropy in the Arrival Direction of Cosmic Rays with KASCADE-Grande,Astrophys

    W.D. Apel et al.,Search for Large-scale Anisotropy in the Arrival Direction of Cosmic Rays with KASCADE-Grande,Astrophys. J.870(2019) 91. [10]IceCubecollaboration,Observation of Cosmic Ray Anisotropy with the IceTop Air Shower Array,Astrophys. J.765(2013) 55 [1210.5278]. [11]IceCubecollaboration,Anisotropy in Cosmic-ray Arrival Directions in the Southern ...

    Pith/arXiv arXiv 2019

  6. [6]

    Hillas,Can diffusive shock acceleration in supernova remnants account for high-energy galactic cosmic rays?,J

    A.M. Hillas,Can diffusive shock acceleration in supernova remnants account for high-energy galactic cosmic rays?,J. Phys. G31(2005) R95

    2005

  7. [7]

    Ahlers,Large- and Medium-Scale Anisotropies in the Arrival Directions of Cosmic Rays observed with KASCADE-Grande,Astrophys

    M. Ahlers,Large- and Medium-Scale Anisotropies in the Arrival Directions of Cosmic Rays observed with KASCADE-Grande,Astrophys. J. Lett.886(2019) L18 [1909.09222]

    Pith/arXiv arXiv 2019

  8. [8]

    Ahlers,Anomalous anisotropies of cosmic rays from turbulent magnetic fields,Phys

    M. Ahlers,Anomalous anisotropies of cosmic rays from turbulent magnetic fields,Phys. Rev. Lett.112(2014) 021101

    2014

  9. [9]

    Kostunin, I

    D. Kostunin, I. Plokhikh, M. Ahlers, V. Tokareva, V. Lenok, P.A. Bezyazeekov et al.,New insights from old cosmic rays: A novel analysis of archival KASCADE data,PoSICRC2021 (2021) 319 [2108.03407]

    Pith/arXiv arXiv 2021

  10. [10]

    J. He, Y. Zhang and Q. Yuan,Observation of large-scale anisotropy of very high-energy cosmic-ray protons with LHAASO-KM2A,PoSICRC2025(2025) 286

    2025

  11. [11]

    Lebedev, A.I

    I.A. Lebedev, A.I. Fedosimova, K.K. Olimov, P.M. Krassovitskiy, N.O. Yerezhep, S.A. Ibraimova et al.,Indications of anisotropy of cosmic-ray elemental groups based on kascade data,The Astrophysical Journal1001(2026) 182. – 20 – [20]KASCADEcollaboration,The Cosmic ray experiment KASCADE,Nucl. Instrum. Meth. A 513(2003) 490

    2026

  12. [12]

    Haungs, D

    A. Haungs, D. Kang, S. Schoo, D. Wochele, J. Wochele, W.D. Apel et al.,The kascade cosmic-ray data centre kcdc: granting open access to astroparticle physics research data,The European Physical Journal C78(2018)

    2018

  13. [13]

    Kuznetsov, N.A

    M.Y. Kuznetsov, N.A. Petrov, I.A. Plokhikh and V.V. Sotnikov,Methods of machine learning for the analysis of cosmic rays mass composition with the KASCADE experiment data,JINST 19(2024) P01025 [2311.06893]

    Pith/arXiv arXiv 2024

  14. [14]

    Kuznetsov, N.A

    M.Y. Kuznetsov, N.A. Petrov, I.A. Plokhikh and V.V. Sotnikov,Energy spectra of elemental groups of cosmic rays with the KASCADE experiment data and machine learning,JCAP05 (2024) 125 [2312.08279]

    Pith/arXiv arXiv 2024

  15. [15]

    Linsley,Fluctuation effects on directional data,Phys

    J. Linsley,Fluctuation effects on directional data,Phys. Rev. Lett.34(1975) 1530

    1975

  16. [16]

    Bonino, V.V

    R. Bonino, V.V. Alekseenko, O. Deligny, P.L. Ghia, M. Grigat, A. Letessier-Selvon et al.,The East-West method: an exposure-independent method to search for large scale anisotropies of cosmic rays,Astrophys. J.738(2011) 67 [1106.2651]

    Pith/arXiv arXiv 2011

  17. [17]

    Peebles,The large-scale structure of the universe(1980)

    P.J.E. Peebles,The large-scale structure of the universe(1980). [27]Pierre Augercollaboration,Arrival Directions of Cosmic Rays above 32 EeV from Phase One of the Pierre Auger Observatory,Astrophys. J.935(2022) 170 [2206.13492]

    Pith/arXiv arXiv 1980

  18. [18]

    Landy and A.S

    S.D. Landy and A.S. Szalay,Bias and Variance of Angular Correlation Functions,Astrophys. J.412(1993) 64

    1993

  19. [19]

    G. Chon, A. Challinor, S. Prunet, E. Hivon and I. Szapudi,Fast estimation of polarization power spectra using correlation functions,Mon. Not. Roy. Astron. Soc.350(2004) 914 [astro-ph/0303414]

    Pith/arXiv arXiv 2004

  20. [20]

    Szapudi, S

    I. Szapudi, S. Prunet, D. Pogosyan, A.S. Szalay and J.R. Bond,Fast Cosmic Microwave Background Analyses via Correlation Functions,Astrophys. J. Lett.548(2001) L115

    2001

  21. [21]

    Sommers,Cosmic ray anisotropy analysis with a full-sky observatory,Astropart

    P. Sommers,Cosmic ray anisotropy analysis with a full-sky observatory,Astropart. Phys.14 (2001) 271 [astro-ph/0004016]

    Pith/arXiv arXiv 2001

  22. [22]

    G´ orski, E

    K.M. G´ orski, E. Hivon, A.J. Banday, B.D. Wandelt, F.K. Hansen, M. Reinecke et al.,Healpix: A framework for high-resolution discretization and fast analysis of data distributed on the sphere,The Astrophysical Journal622(2005) 759

    2005

  23. [23]

    Zonca, L.P

    A. Zonca, L.P. Singer, D. Lenz, M. Reinecke, C. Rosset, E. Hivon et al.,healpy: equal area pixelization and spherical harmonics transforms for data on the sphere in python,Journal of Open Source Software4(2019) 1298

    2019

  24. [24]

    Wochele, D

    J. Wochele, D. Kang, D. Wochele, A. Haungs and S. Schoo,Kcdc user manual: Open access solution for the kascade, Tech. Rep. https://doi.org/10.17616/R3TS4P, KASCADE Cosmic-ray Data Centre (KCDC) (2013)

    work page doi:10.17616/r3ts4p 2013

  25. [25]

    Alexandreas, D

    D. Alexandreas, D. Berley, S. Biller, G. Dion, J. Goodman, T. Haines et al.,Point source search techniques in ultra high energy gamma ray astronomy,Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment328(1993) 570

    1993

  26. [26]

    Antoni, W.D

    T. Antoni, W.D. Apel, A.F. Badea, K. Bekk, A. Bercuci, H. Bl¨ umer et al.,Large-scale cosmic-ray anisotropy with kascade,The Astrophysical Journal604(2004) 687. – 21 –

    2004