Pith. sign in

REVIEW 2 major objections 4 minor 43 references

The Hellings-Downs correlation, the angular fingerprint of a gravitational-wave background in pulsar timing arrays, has a skewness that does not vanish as the number of sources grows to infinity.

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 · deepseek-v4-flash

2026-08-03 05:51 UTC pith:XZBAOSIJ

load-bearing objection Solid third-moment extension of Allen's HD variance, but the O(1) skewness is a single-frequency-bin result that likely shrinks when summed over a PTA band. the 2 major comments →

arxiv 2602.01108 v2 pith:XZBAOSIJ submitted 2026-02-01 astro-ph.CO

Skewness in the Hellings-Downs curve

Ryosuke Fujimoto , Keitaro Takahashi This is my paper
classification astro-ph.CO
keywords pulsar timing arraysHellings-Downs correlationstochastic gravitational wave backgroundskewnessnon-Gaussianitycosmic variancethree-point correlationsource discreteness
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper asks whether the Hellings-Downs correlation—the expected angular correlation between pulsar timing residuals produced by a stochastic gravitational-wave background—is Gaussian. The authors compute its third central moment, the skewness, for a single point source and for a superposition of many interfering discrete sources, and find that the skewness does not vanish in the limit of infinitely many sources. In that limit the normalized skewness is a nonzero, order-one function of pulsar separation whose sign everywhere matches the mean correlation. By pulsar-averaging before forming moments, they isolate a 'cosmic' skewness controlled by a newly introduced three-point average function, showing that the background is intrinsically non-Gaussian even after all pulsar-dependent noise is removed. If this carries to realistic source populations, skewness becomes a new observable probe of source discreteness beyond the mean HD curve and its variance.

Core claim

The paper derives the skewness of the Hellings-Downs correlation in the many-source interference regime and finds it does not vanish as the number of sources grows: the leading large-N term is κ₃ ≈ (1/16)H₂³[µ_u³(γ) + 12µ_u(γ)µ_u²(0)], which after normalization gives an O(1) skewness S(γ) whose sign everywhere matches the HD mean. Pulsar-averaging yields the cosmic skewness, S_cosmic → 2 cµ₃(γ)/(fµ₂(γ))^{3/2}, governed by a new three-point average cµ₃(γ)=⟨µ(γ,β₁₂)µ(γ,β₁₃)µ(γ,β₂₃)⟩ over source directions.

What carries the argument

The argument is carried by the interfering-source model: all N sources radiate at the same angular frequency with independent uniform phases, so averaging over phases leaves only terms where source indices coincide (formally ⟨e^{i(φ_j−φ_k)}⟩ = δ_jk). This index contraction produces the amplitude-sum factors H₂, H₄, H₆, and in the large-N limit the skewness is dominated by the H₂³ term. A second key object is the three-point-average function cµ₃(γ), which isolates the cosmic skewness after pulsar-averaging.

Load-bearing premise

The load-bearing premise is that every source radiates at exactly the same frequency with independent random phases; if in reality the sources occupy different frequency bins and lose phase coherence, the calculated order-one skewness could shrink or vanish.

What would settle it

Repeat the third-moment derivation for a model with sources at two distinct frequencies separated by more than the inverse observation time, and check whether the normalized skewness tends to zero as the second-frequency fraction approaches one; if it does, the O(1) result is an artifact of the single-frequency interfering-source model.

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

If this is right

  • The probability distribution of the HD correlation is asymmetric with |mode| < |median| < |mean|, so most measured correlations are closer to zero than the mean, with rare large-correlation events pulling the mean.
  • The cosmic skewness is nonzero for N ≥ 3 and is governed by the new three-point function cµ₃(γ), so third-order statistics carry source-discreteness information that Gaussian analyses lack.
  • Because the normalized skewness is independent of the total intensity H₂, it can be compared across datasets and frequency bands without knowing the absolute background amplitude.
  • The order-unity skewness suggests a hierarchical Bayesian model in which the overlap reduction function is treated as a non-Gaussian random variable with mean anchored to HD and fluctuations informed by cosmic variance and skewness.
  • Skewness alone is an order-one effect, so kurtosis and higher cumulants would be needed to describe the distribution tails, especially for small numbers of bright sources.

Where Pith is reading between the lines

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

  • If the equal-frequency assumption is relaxed, inter-source phase coherence may break for sources in different frequency bins, and the O(1) skewness could shrink or vanish; computing the third moment for a two-frequency population would show whether Eq. (3.40) is an upper bound or a generic prediction.
  • The three-point function cµ₃(γ) could be recomputed for anisotropic source distributions or different polarization content, turning cosmic skewness into a possible discriminator between astrophysical and exotic background models.
  • A finite-pulsar-distance calculation would reveal how strongly the constant 12µ_u²(0) depends on the pulsar-term parameter χ; that dependence is currently absorbed by the large-distance averaging assumption.
  • Because the sign of the total skewness is always the same as the HD mean, one could build a one-sided test using only pulsar pairs on the positive (or negative) side of the curve, increasing sensitivity while keeping the mean signal near zero.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper computes the third central moment (skewness) of the inter-pulsar Hellings–Downs (HD) correlation for two models: a single unpolarized point source and an ensemble of N monochromatic interfering sources. It also introduces a 'three-point average' to define a pulsar-averaged cosmic skewness. The central claimed results are Eq. (3.42), a normalized many-source skewness that is O(1) and proportional to the HD mean in the large-N limit, and Eq. (4.15), a cosmic skewness controlled by the new three-point function cµ3(γ). The internal algebra, including the cancellations in Eqs. (3.37)–(3.40) and (4.7)–(4.12), is consistent.

Significance. If the frequency-integration issue were resolved, this would be a useful extension of Allen (2023), showing that source discreteness can leave a non-Gaussian signature even in the large-source-number limit. The paper is transparent in its derivations, and the large-N statements are correct within the stated model. However, the physical significance is conditional: the model assumes all sources share a single frequency, and the paper does not connect this to a realistic multi-frequency PTA analysis, where the normalized skewness of the summed correlation is suppressed. This limits the strength of the observational conclusion as currently presented.

major comments (2)
  1. [Sec. III, Eq. (3.7) and Eq. (3.42)] The phase-coherence property that generates the nonzero skewness requires identical frequency for all N sources. In a PTA, sources radiate over a band; after time-averaging over observation T, pairs separated by Δω ≫ 1/T are phase-incoherent, so the total correlation is effectively a sum over M independent frequency bins. With comparable per-bin intensity, κ3,total = M κ3,bin and σ²_total = M σ²_bin, so S_total = S_bin/√M → 0 as M grows. The paper does not compute the band-integrated third moment; Sec. V only writes N(f) as a parameter. Without this computation, the claimed O(1) skewness in Eq. (3.42) is an artifact of the single-frequency idealization. Please either provide the multi-bin calculation or explicitly limit the claim to one bin with a quantitative statement of the suppression.
  2. [Sec. IV C, Eq. (4.15)] The cosmic skewness has the same frequency-averaging problem. The three-point function cµ3(γ) is a single-frequency quantity. In a realistic search, independent frequency bins add incoherently, so the normalized cosmic skewness of the total also carries a 1/√M factor. The proposed hierarchical model in Eq. (5.2) is therefore not yet connected to an observable; the paper should either show how per-bin skewness enters a band-integrated likelihood or weaken the Discussion claims.
minor comments (4)
  1. [Sec. II C, Eq. (2.9)] The expression for ⟨ρ_u³⟩ is stated without derivation. Since it is presented as a new result rather than a direct quote of Allen (2023), include the integration steps or an appendix; as written the reader cannot verify it.
  2. [Sec. V, Eq. (5.2)] The hierarchical prior p(Γ|N) is schematic. The text does not specify how the computed cµ3(γ) and fµ2(γ) would be used to construct the prior, so the practical impact is unclear.
  3. [Fig. 7 caption] The caption calls the plotted quantity the 'amplitude of cosmic skewness,' but it is the coefficient in Eq. (4.14). Consider renaming it to 'cosmic skewness coefficient' to avoid ambiguity.
  4. [Sec. III, Eq. (3.1)] The deterministic amplitude law A_j = j^{−1/3}A is an idealized model. The finite-N behavior in Fig. 7 depends on this law, so the dependence should be stated as a caveat when using Fig. 7 to discuss constraints on N.

Circularity Check

0 steps flagged

No significant circularity: the skewness formulas are explicit ensemble averages with no fitted inputs or self-citations.

full rationale

The central results are derived by direct ensemble averaging over source phases and directions. Eq. (3.40) follows by subtracting the explicitly computed mean (3.19) and variance (3.30) from the computed third moment (3.37); every term is exhibited, and no parameter is fitted to the quantity being predicted. The normalized skewness (3.42) inherits the cancellation of H2 factors from this computation, not from a definition. The cosmic skewness in Eq. (4.12) is obtained after exact cancellation of the 1/8 H2^3 mu_u^3 and 3/8 H2(H2^2-H4) mu_u fmu2 terms; although the new three-point function c_mu3 is introduced as the remaining averaged object, it is defined independently in Eq. (4.11) and then evaluated numerically in Appendix B, so the result is a genuine algebraic reduction rather than a renaming. The cited formulas from Allen (2023) and Allen & Romano (2023) are prior external work by different authors, not self-citations, and they supply input ingredients (single-source variance and the two-point function) rather than forcing the new third-moment result. The single-frequency, identical-omega assumption is an explicit physical modeling assumption, not an output smuggled in by definition or by a fitting procedure. Consequently, no circular step satisfying the required quote-and-reduction standard can be identified.

Axiom & Free-Parameter Ledger

3 free parameters · 7 axioms · 1 invented entities

The calculation rests on the Allen (2023) confusion-noise model: identical-frequency unpolarized sources with independent phases and deterministic j^{-1/3} amplitudes, plus the large-pulsar-distance approximation for pulsar-term phases. None of these are fitted to data; they are the model assumptions that define the problem. The only genuinely new input object is the three-point function cµ₃(γ), which is derived from prior building blocks rather than postulated.

free parameters (3)
  • N (number of sources)
    Effective number of monochromatic sources in the model; enters H₂, H₄, H₆ and the finite-N skewness amplitude (Fig. 7), but is not fitted to data; the paper only maps it to an observable schematically.
  • χ (pulsar-term reflection parameter) = 1 (set by hand)
    Set to 1 ('if we properly include the pulsar term') in the main results; the constants 4 and 12 in Eqs. (3.30)/(3.40) follow from (1+χ²)² at χ=1. The value is a modeling choice, not derived or fitted.
  • A (nearest-source amplitude)
    Overall strain scale of the nearest source; cancels in all normalized results and is a physical input, not fitted.
axioms (7)
  • domain assumption Source phases φ_j are statistically independent and uniform on [0, 2π)
    Sec. III, Eq. (3.8); defines the ensemble and yields the δ_jk phase correlations that generate all interference terms carrying the skewness.
  • domain assumption Uniform spatial distribution with deterministic amplitudes A_j = j^{-1/3} A
    Sec. III, Eq. (3.1); the authors explicitly note it could be a probability distribution but adopt the deterministic form 'to enable analytic discussion'. It controls the H₂ ∝ N^{1/3} scaling and the finite-N behavior of Fig. 7.
  • domain assumption Rapidly oscillating pulsar-term phases average to zero over source directions: ⟨B₁B₂*⟩ ≈ 1 and ⟨|B|²⟩ ≈ 1 + χ²
    Eqs. (3.17) and (3.26); requires large pulsar distances L₁, L₂. Fixes the µ_u(0) factors and hence the 12µ²_u(0) constant in Eq. (3.40).
  • domain assumption All N sources radiate at the same angular frequency ω (interfering / confusion-noise model)
    Sec. III opening; without identical frequencies, the phase-interference terms that carry the skewness would not survive time-averaging.
  • domain assumption Two-point function µ(γ,β) and single-source moments σ²_u, σ²_p, σ²_c taken from Allen (2023) and Allen & Romano (2023)
    Eqs. (B.1) and (2.4)–(2.7); imported results are the building blocks of the three-point function and of the cancellation structure.
  • domain assumption Dominant-term truncation: H₂³ ≫ H₄H₂ and H₂³ ≫ H₆ for large N
    Sec. III C and Appendix A; justified by H₂ ∝ N^{1/3}, H₄, H₆ → const, but subdominant terms are never computed, so the exact finite-N skewness is unknown.
  • standard math Random-phase moment combinatorics: ⟨e^{i(φ_a−φ_b+...)}⟩ evaluated via Kronecker-delta index pairings
    Used in Eqs. (3.22), (A.6); standard evaluation of moments of uniform-phase exponentials.
invented entities (1)
  • three-point-average function cµ₃(γ) = ⟨µ(γ,β₁₂)µ(γ,β₁₃)µ(γ,β₂₃)⟩_Ω independent evidence
    purpose: Encodes the connected three-source contribution to the cosmic skewness; it is the coefficient of (H₂³ − 3H₄H₂ + 2H₆) in κ₃,cosmic (Eq. 4.12) and controls S_cosmic in the large-N limit (Eq. 4.15).
    A new statistical object, but defined entirely from the previously known two-point function µ(γ,β); it is computed by a specified numerical integral (Appendix B), so it is externally checkable and parameter-free. It is a geometric correlation average, not a new physical entity.

pith-pipeline@v1.3.0-alltime-deepseek · 18704 in / 41753 out tokens · 410309 ms · 2026-08-03T05:51:16.123312+00:00 · methodology

0 comments
read the original abstract

Recent Pulsar Timing Array datasets provide compelling evidence for a nano-Hertz gravitational-wave background, but robust detection requires characterizing statistical fluctuations of the Hellings-Downs (HD) correlation expected from a finite population of discrete sources. Building on the variance calculation of Allen (2023), we derive the third central moment (skewness) of the HD correlation for a single unpolarized point source and an ensemble of many interfering point sources in the confusion-noise regime. To isolate the intrinsic non-Gaussianity of the background, we extend the pulsar-averaging formalism to third order by introducing a three-point averaged correlation function, which allows us to define the cosmic skewness. We find that the skewness remains non-zero in the large-source-number limit and is controlled by a new geometric three-point function. These results suggest that incorporating higher-order moments could provide additional information on source discreteness beyond standard Gaussian analyses.

Figures

Figures reproduced from arXiv: 2602.01108 by Keitaro Takahashi, Ryosuke Fujimoto.

Figure 1
Figure 1. Figure 1: FIG. 1. Comparison of single-source statistical quantities. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Statistical quantities for a large number of sources [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. The many-point skewness [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. The amplitude of cosmic skewness, defined as 2( [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. The cosmic skewness [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. The direction of the first GW source is along the [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 6 linked inside Pith

  1. [1]

    III, the⟨ρ 3 diag⟩term is independent of phase

    The⟨ρ 3 diag⟩Term As shown in Sec. III, the⟨ρ 3 diag⟩term is independent of phase. The expansion contains terms proportional toH 6,H 4H2, andH 3

  2. [2]

    In the large-Nlimit, thei̸=j̸=kterm dominates. Performing the direction average on this dominant term yields: ⟨ρ3 diag⟩ ≈ * X i̸=j̸=k (cid∗ i +c ∗ i di)(cjd∗ j +c ∗ j dj)(ckd∗ k +c ∗ kdk) + Ω = 1 8 H3 2µ3 u(γ).(3.34) 2.The⟨3ρ diagρ2 off-diag⟩T erm Due to the complexity of this calculation, the de- tailed steps are presented in Appendix A. The dom- 7 inant...

    2023

  3. [3]

    This indicates that the SGWB formed by interfering discrete sources possesses an intrin- sic non-Gaussian nature that persists even in the high- source-count regime

    Consequently, the normalized skewness con- verges to a non-zero function that becomes independent of the intensity scale. This indicates that the SGWB formed by interfering discrete sources possesses an intrin- sic non-Gaussian nature that persists even in the high- source-count regime. Furthermore, we successfully isolated the ”Cosmic Skewness” by removi...

  4. [4]

    In the large-Nlimit, the contribution from thei̸=j̸=kterms becomes dominant. Consequently, the ensemble average yields: ⟨ρ3 diag⟩= *X i (cid∗ i +c ∗ i di) +3 Ω = *X i (cid∗ i +c ∗ i di)3 + Ω + 3 *X i̸=j (cid∗ i +c ∗ i di)(cjd∗ j +c ∗ j dj)2 + Ω + * X i̸=j̸=k (cid∗ i +c ∗ i di)(cjd∗ j +c ∗ j dj)(ckd∗ k +c ∗ kdk) + Ω ≈ 1 8 H3 2µ3 u(γ),(A.1) where P i(cid∗ i...

  5. [5]

    X j̸=k n cjd∗ kei(ϕj −ϕk) +c ∗ j dke−i(ϕj −ϕk) o #

    The⟨3ρ diagρ2 off-diag⟩Term 3ρdiagρ2 off-diag = * 3 X i (cid∗ i +c ∗ i di) "X j̸=k n cjd∗ kei(ϕj −ϕk) +c ∗ j dke−i(ϕj −ϕk) o #" X ℓ̸=m n cjd∗ kei(ϕℓ−ϕm) +c ∗ ℓ dme−i(ϕℓ−ϕm) o #+ = * 3 X i (cid∗ i +c ∗ i di) X j̸=k X ℓ̸=m h cjd∗ kcℓd∗ m⟨ei(ϕj −ϕk+ϕℓ−ϕm)⟩ϕ +c jd∗ kc∗ ℓ dm⟨ei(ϕj −ϕk−ϕℓ+ϕm)⟩ϕ+ c∗ j dkc∗ ℓ dm⟨e−i(ϕj −ϕk+ϕℓ−ϕm)⟩ϕ +c ∗ j dkcℓd∗ m⟨e−i(ϕj −ϕk−ϕℓ+ϕ...

  6. [6]

    The calculation requires evaluating six-phase correlation, such as⟨e i(ϕi−ϕj +ϕk−ϕℓ+ϕm−ϕn)⟩ϕ

    The⟨ρ 3 off-diag⟩Term This term represents the cubic self-interaction of the off-diagonal interference terms, written as: ⟨ρ3 off-diag⟩ϕ = *X i̸=j X k̸=ℓ X m̸=n h cid∗ j ei(ϕi−ϕj ) +c ∗ i dje−i(ϕi−ϕj ) ih ckd∗ ℓ ei(ϕk−ϕℓ) +c ∗ kdℓe−i(ϕk−ϕℓ) i × h cmd∗ nei(ϕm−ϕn) +c ∗ mdne−i(ϕm−ϕn) i+ ϕ = X i̸=j X k̸=ℓ X m̸=n h cid∗ j ckd∗ ℓ cmd∗ n⟨ei(ϕi−ϕj +ϕk−ϕℓ+ϕm−ϕn)⟩ϕ...

    2023

  7. [7]

    Detweiler, Pulsar timing measurements and the search for gravitational waves, Astrophys

    S. Detweiler, Pulsar timing measurements and the search for gravitational waves, Astrophys. J.234, 1100 (1979)

    1979

  8. [8]

    R. S. Foster and D. C. Backer, Constructing a pulsar timing array, Astrophys. J.361, 300 (1990)

    1990

  9. [9]

    Burke-Spolaoret al., The astrophysics of nanohertz gravitational waves, Astron

    S. Burke-Spolaoret al., The astrophysics of nanohertz gravitational waves, Astron. Astrophys. Rev.27, 5 (2019)

    2019

  10. [10]

    J. P. W. Verbiestet al.(IPTA Collaboration), The in- ternational pulsar timing array: First data release, Mon. Not. R. Astron. Soc.458, 1267 (2016)

    2016

  11. [11]

    M. V. Sazhin, Opportunities for detecting ultralong grav- itational waves, Sov. Astron.22, 36 (1978)

    1978

  12. [12]

    Agazieet al.(NANOGrav Collaboration), The NANOGrav 15 yr data set: Evidence for a gravitational- wave background, Astrophys

    G. Agazieet al.(NANOGrav Collaboration), The NANOGrav 15 yr data set: Evidence for a gravitational- wave background, Astrophys. J. Lett.951, L8 (2023)

    2023

  13. [13]

    A. D. Johnsonet al.(NANOGrav Collaboration), The NANOGrav 15-year Gravitational-Wave Background Methods, Physical Review D109, 103012 (2024), 15 arXiv:2306.16223 [astro-ph.HE]

    Pith/arXiv arXiv 2024

  14. [14]

    Antoniadiset al.(EPTA Collaboration and InPTA Collaboration), The second data release from the Euro- pean pulsar timing array: III

    J. Antoniadiset al.(EPTA Collaboration and InPTA Collaboration), The second data release from the Euro- pean pulsar timing array: III. search for gravitational wave signals, Astron. Astrophys.678, A50 (2023)

    2023

  15. [15]

    D. J. Reardonet al.(PPTA Collaboration), Search for an isotropic gravitational-wave background with the Parkes pulsar timing array, Astrophys. J. Lett.951, L6 (2023)

    2023

  16. [16]

    Xuet al.(CPTA Collaboration), Searching for the nano-hertz stochastic gravitational wave background with the Chinese pulsar timing array data release I, Res

    H. Xuet al.(CPTA Collaboration), Searching for the nano-hertz stochastic gravitational wave background with the Chinese pulsar timing array data release I, Res. Astron. Astrophys.23, 075024 (2023)

    2023

  17. [17]

    Raviet al., Does a ”stochastic” background of gravita- tional waves exist in the pulsar timing band?, Astrophys

    V. Raviet al., Does a ”stochastic” background of gravita- tional waves exist in the pulsar timing band?, Astrophys. J.761, 84 (2012)

    2012

  18. [18]

    Babak and A

    S. Babak and A. Sesana, Resolving multiple supermassive black hole binaries with pulsar timing arrays, Phys. Rev. D85, 044034 (2012)

    2012

  19. [19]

    R. M. Shannonet al., Gravitational waves from binary supermassive black holes: Missing in the Parkes pulsar timing array, Science349, 1522 (2015)

    2015

  20. [20]

    E. S. Phinney, A practical theorem on gravitational wave backgrounds, arXiv preprint astro-ph/0108028 10.48550/arXiv.astro-ph/0108028 (2001)

    Pith/arXiv arXiv doi:10.48550/arxiv.astro-ph/0108028 2001

  21. [21]

    A. H. Jaffe and D. C. Backer, Gravitational waves from supermassive black hole mergers in the era of LISA and pulsar timing arrays, Astrophys. J.583, 616 (2003)

    2003

  22. [22]

    Sesanaet al., The stochastic gravitational wave back- ground from massive black hole binary populations, Mon

    A. Sesanaet al., The stochastic gravitational wave back- ground from massive black hole binary populations, Mon. Not. R. Astron. Soc.390, 192 (2008)

    2008

  23. [23]

    Sesana, Systematic investigation of the expected grav- itational wave signal from supermassive black hole bina- ries in the pulsar timing band, Mon

    A. Sesana, Systematic investigation of the expected grav- itational wave signal from supermassive black hole bina- ries in the pulsar timing band, Mon. Not. R. Astron. Soc. 433, L1 (2013)

    2013

  24. [24]

    P. A. Rosado, A. Sesana, and J. Gair, Expected proper- ties of the gravitational wave signal from supermassive black hole binaries, Mon. Not. R. Astron. Soc.451, 2417 (2015)

    2015

  25. [25]

    Christensen, Stochastic gravitational wave back- grounds, Rep

    N. Christensen, Stochastic gravitational wave back- grounds, Rep. Prog. Phys.82, 016903 (2018)

    2018

  26. [26]

    R. W. Hellings and G. S. Downs, Upper limits on the isotropic gravitational radiation background from pulsar timing analysis, Astrophys. J.265, L39 (1983)

    1983

  27. [27]

    F. A. Jenet, G. B. Hobbs, K. J. Lee, and R. N. Manch- ester, Detecting the stochastic gravitational wave back- ground using pulsar timing, Astrophys. J. Lett.625, L123 (2005)

    2005

  28. [28]

    Roebberet al., Cosmic variance in the nanohertz grav- itational wave background, Astrophys

    E. Roebberet al., Cosmic variance in the nanohertz grav- itational wave background, Astrophys. J.831, 53 (2016)

    2016

  29. [29]

    Allen, Variance of the Hellings-Downs correlation, Phys

    B. Allen, Variance of the Hellings-Downs correlation, Phys. Rev. D107, 043018 (2023)

    2023

  30. [30]

    Allen and J

    B. Allen and J. D. Romano, Hellings and Downs corre- lation of an arbitrary set of pulsars, Phys. Rev. D107, 043513 (2023)

    2023

  31. [31]

    Anholmet al., Optimal strategies for gravitational wave stochastic background searches in pulsar timing data, Phys

    M. Anholmet al., Optimal strategies for gravitational wave stochastic background searches in pulsar timing data, Phys. Rev. D79, 084030 (2009)

    2009

  32. [32]

    Maggiore, Gravitational wave experiments and early universe cosmology, Phys

    M. Maggiore, Gravitational wave experiments and early universe cosmology, Phys. Rep.331, 283 (2000)

    2000

  33. [33]

    N. J. Cornish and A. Sesana, Pulsar timing array analy- sis for black hole backgrounds, Phys. Rev. D87, 044007 (2013)

    2013

  34. [34]

    J. D. Romano and N. J. Cornish, Detection methods for stochastic gravitational-wave backgrounds: a unified treatment, Living Rev. Relativ.20, 2 (2017)

    2017

  35. [35]

    Bartoloet al., Probing non-Gaussianities in the cos- mological gravitational-wave background with LISA, J

    N. Bartoloet al., Probing non-Gaussianities in the cos- mological gravitational-wave background with LISA, J. Cosmol. Astropart. Phys.2018, 034

    2018

  36. [36]

    Sesana, A

    A. Sesana, A. Vecchio, and M. Volonteri, Gravitational waves from massive black hole binaries in the lisa window, Mon. Not. R. Astron. Soc.394, 2255 (2009)

    2009

  37. [37]

    B´ ecsyet al., Exploring realistic nanohertz gravitational-wave backgrounds, Astrophys

    B. B´ ecsyet al., Exploring realistic nanohertz gravitational-wave backgrounds, Astrophys. J.936, 140 (2022)

    2022

  38. [38]

    S. R. Taylor and J. R. Gair, Searching for anisotropic gravitational-wave backgrounds using pulsar timing ar- rays, Phys. Rev. D88, 084001 (2013)

    2013

  39. [39]

    C. M. F. Mingarelliet al., Characterizing gravitational wave stochastic background anisotropy with pulsar tim- ing arrays, Phys. Rev. D88, 062005 (2013)

    2013

  40. [40]

    N. J. Cornish and R. van Haasteren, Mapping the nanohertz gravitational wave sky, arXiv preprint arXiv:1406.4511 (2014)

    Pith/arXiv arXiv 2014

  41. [41]

    G. Agazieet al.(NANOGrav Collaboration), The NANOGrav 15-year Data Set: Search for Anisotropy in the Gravitational-Wave Background, The Astrophysical Journal Letters956, L3 (2023), arXiv:2306.16221 [astro- ph.HE]

    Pith/arXiv arXiv 2023

  42. [42]

    G. Agazieet al.(NANOGrav Collaboration), The NANOGrav 15-year Data Set: Search for Transverse Po- larization Modes in the Gravitational-Wave Background, The Astrophysical Journal Letters964, L14 (2024), arXiv:2310.12138 [gr-qc]

    Pith/arXiv arXiv 2024

  43. [43]

    R. C. Bernardo and K.-W. Ng, Hunting the stochastic gravitational wave background in pulsar timing array cross correlations through theoretical uncertainty, JCAP 08, 028, arXiv:2304.07040 [gr-qc]

    Pith/arXiv arXiv