Pith. sign in

REVIEW 2 major objections 5 minor 38 references

On Stein's test of uniformity on the hypersphere

T0 review · 2 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read This paper proposes a new uniformity test on the hypersphere, proves it is consistent against all absolutely continuous non-uniform distributions, and derives its explicit asymptotic distribution.

desk verdict A solid, genuinely new Sobolev-class uniformity test with careful asymptotics; the empirical power story leans on oracle tuning and the claimed omnibus consistency exceeds the L2-density proof. read the letter →

arxiv 2602.20896 v2 pith:3ZLFHRSY submitted 2026-02-24 math.ST stat.MEstat.TH

Paul Axmann , Bruno Ebner , Eduardo García-Portugués This is my paper · ORCID
classification math.STstat.MEstat.TH MSC 62H1162G1062G20
keywords directionalstatisticsuniformitytestSteincharacterizationLaplace-BeltramioperatorsphericalharmonicsGegenbauerpolynomialsSobolevteststuningparameter
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 builds a test of uniformity on the sphere S^{p-1} from a Stein characterization tied to the Laplace–Beltrami operator. The test checks whether the empirical average of Δ(e^{λ t·X}) vanishes in L² over all directions t, and the authors prove that only the uniform law makes the population version zero. They show the statistic has a closed-form Gegenbauer expansion, belongs to the classical Sobolev class of tests, and converges under the null to an explicit weighted sum of chi-squares. Under fixed alternatives the statistic grows like n times a positive constant, so the test detects every absolutely continuous non-uniform distribution. A tuning parameter λ shifts sensitivity from low-order to high-order features, with data-driven selection that improves power against concentrated and multimodal alternatives.

What carries the argument

The central object is the statistic T_n(λ) = ‖ n^{-1/2} Σ_j Δ_{S^{p-1}} e^{λ t·X_j} ‖²_{L²(S^{p-1})}. The mechanism that carries the argument is the eigenfunction relation for the Laplace–Beltrami operator: Gegenbauer (and Chebyshev) polynomials are eigenfunctions, so the exponential test function e^{λ t·x} has an explicit harmonic expansion with coefficients involving modified Bessel functions. This turns the statistic into a double sum over Gegenbauer polynomials of X_i·X_j, reveals it as a Sobolev test, and yields closed-form series for the null distribution, the fixed-alternative limit, and the limiting covariance kernels.

What would settle it

Simulate U(0,1)-distributed samples from the uniform law on S², compute T_n(λ) for λ = 1 and n = 1000 over 100,000 replicates, and compare the empirical 95th percentile to the closed-form null series Σ_k c_{k,3}(1) γ_{k,3} χ²_{2k+1} truncated at k = 100. A systematic mismatch beyond Monte Carlo error would refute the claimed null distribution.

Watch

Extended reading notes

Core claim

For an absolutely continuous random vector X on S^{p-1}, the identity E[Δ_{S^{p-1}} e^{λ t·X}] = 0 for every t ∈ S^{p-1} holds if and only if X is uniform. The paper therefore defines T_n(λ), the squared L² norm of the empirical version of that quantity, and proves its asymptotic theory: under H0 it converges to Σ_{k=1}∞ c_{k,p}(λ) γ_{k,p} χ²_{d_{k,p}}, and under fixed alternatives T_n(λ)/n converges almost surely to a positive constant, giving consistency. The construction is a Sobolev test with positive weights on every spherical-harmonic degree, and the tuning parameter λ interpolates between a low-order moment test and a maximum inner-product-type test.

Load-bearing premise

The consistency and characterization arguments assume every alternative distribution has a square-integrable density with respect to the uniform measure on the sphere; if an alternative has atoms or a very rough density, the spherical-harmonic expansion and the proof that τ > 0 can fail.

Editorial extensions

If this is right

  • The test provides an omnibus, consistent procedure for uniformity on S^{p-1}: any absolutely continuous non-uniform distribution is eventually detected with probability one.
  • The closed-form null distribution allows asymptotic p-values and power curves to be computed without simulation, once the series is truncated at a sufficiently large order.
  • The tuning parameter λ gives a principled way to trade off detection of diffuse low-frequency deviations against concentrated or multimodal deviations, with a data-driven selection rule.
  • Because the statistic is a Sobolev test with positive weights at every degree, it connects the Stein-characterization route to the classical Beran–Giné framework of rotation-invariant tests.
  • For rotationally symmetric alternatives such as the von Mises–Fisher distribution, both the limiting signal τ and the asymptotic variance have explicit expressions, enabling direct power comparisons.

Reading between the lines

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

  • Allowing λ to be complex would replace the moment-generating function with the characteristic function and the modified Bessel coefficients with ordinary Bessel coefficients; the paper notes this possibility, and it could yield a family of oscillatory kernels with different detection profiles for asymmetric alternatives.
  • The data-driven tuning rule is demonstrated with an oracle sample from the alternative; in practice a cross-validated version will generally achieve less than the oracle power, so the reported gains are an upper bound for what is attainable from data alone.
  • The construction relies only on having Laplace–Beltrami eigenfunctions, so it extends in principle to other compact manifolds with empty boundary, though the explicit Gegenbauer coefficients and Funk–Hecke simplifications would be lost and numerical eigenfunctions would be needed.
  • The λ→∞ limit reduces the statistic to a maximum over pairwise inner products, which suggests the test could be particularly strong against very concentrated or sparse alternatives, and connects to high-dimensional phenomena of random angles on spheres.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes a new test of uniformity on the hypersphere S^{p-1} based on the Laplace–Beltrami Stein operator applied to the exponential test functions e^{λt·x}. The statistic is T_n(λ)=||n^{-1/2}∑_j Δ_{S^{p-1}}e^{λt·X_j}||^2_{L^2(S^{p-1})}. The authors derive an explicit Gegenbauer/spherical-harmonic expansion of T_n(λ), show that it belongs to the class of Sobolev tests, and obtain: the null limit as a weighted series of independent chi-square variables (Theorem 3.2), an almost-sure limit under fixed alternatives (Theorem 3.3), and asymptotic normality of the centered statistic (Theorem 3.5). They also analyze the limits λ→0 and λ→∞, connect the construction to the dKSD test, propose a data-driven tuning procedure, and report extensive Monte Carlo power comparisons.

Significance. If the results hold, the paper is a valuable contribution to directional statistics. It adds a new member to the Sobolev class of uniformity tests with explicit, easily computable coefficients; closed-form null and alternative asymptotic distributions; a tuning parameter that interpolates between Rayleigh-type and extreme-angle-type behavior; and a clean spectral relation to the dKSD test. The derivations are careful and internally consistent: the coefficient identities from Bessel integrals, the Funk–Hecke steps, the Karhunen–Loève expansion, and the variance formulas all check out. The main caveats concern the scope of the consistency theorem (square-integrable densities only) and the implementability of the adaptive tuning procedure; both are repairable but currently weaken the strength of some claims.

major comments (2)
  1. [Proposition 1.1, Theorem 3.3, Remark 3.1, Appendix A] The consistency claim is formally proved only for densities q∈L^2(S^{p-1}), while Proposition 1.1 and Remark 3.1 assert characterization and consistency for all absolutely continuous distributions. The proof of Proposition 1.1 begins by assuming q∈L^2 and uses the L^2 harmonic expansion (10); Lemma 3.1 and Theorem 3.3 carry the same q∈L^2 assumption. Many absolutely continuous non-uniform laws on S^{p-1} have densities that are integrable but not square-integrable (e.g., q(x)∝||x−e_1||^{−β} for (p−1)/2<β<p−1). Thus Remark 3.1's statement that Theorem 3.3 implies consistency against all absolutely continuous non-uniform distributions is unsupported as written. The gap is likely repairable: z(s)=Δ_S M_X(λs)=0 implies M_X(λs) is constant on S^{p-1}, and strict positivity of m_{k,p}(λ) together with the addition theorem should force all nonconstant harmonic moments of P to vanish, without re
  2. [Sections 5.2–5.3, Tables 1–4] The adaptive tuning procedure is evaluated largely with an oracle version. The statistic T_n(λ̃) defined in (20) uses 10,000 independent draws from the candidate alternative density to estimate E[A_k]; this information is unavailable in practice. The paper mentions cross-validation as the practical alternative, but the cross-validated statistic T_{n,20}(λ) reported in Table 1 is not covered by the theoretical null distribution in Theorem 3.2, and no size calibration or asymptotic theory is given for it. Consequently, the headline power comparisons in Tables 2–4 for T_n(λ̃) are oracle-based and do not by themselves support the abstract's claim of a data-driven strategy. The authors should either (i) present the cross-validated procedure as the main tuned test with a valid null calibration, or (ii) clearly label T_n(λ̃) as an oracle benchmark and present the implementable variant separatel
minor comments (5)
  1. [Theorem 3.5] The statement contains a duplicated 'S^{p-1} S^{p-1}': 'random vector X on S^{p-1} S^{p-1} with density q'.
  2. [Theorem 3.4] In the displayed covariance kernel, the second term is written as '−Δ_{S^{p-1}}M_X(λs)Δ_{S^{p-1}}M_X(λt)'; this should be z(s)z(t) with the operator acting appropriately on s and t. The current notation is ambiguous.
  3. [Example 3.1] There is a typo: 'explicitly derive the coefficients βk in closed form for for the von Mises–Fisher' — delete the duplicated 'for'.
  4. [Section 5.2] The grid for λ̃ is {i/10 : i=1,...,300}, but Figure 5 plots λ up to 20. It would help to state the grid or range used in Figure 5.
  5. [Section 4.2] The representation of general Sobolev tests as L^2-Stein tests assumes the coefficients b_{k,p} are such that the defining series for f_t converges in L^2(S^{p-1}); this integrability condition is not stated.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the test statistic, null distribution, and fixed-alternative limits are derived from standard harmonic analysis and Hilbert-space asymptotics; λ is a user tuning parameter and no load-bearing self-citation appears.

full rationale

The central derivation chain is self-contained. Proposition 1.1 is proved in Appendix A by expanding the density q in spherical harmonics and using Funk–Hecke; it does not invoke the test statistic or any fitted constant. Lemma 2.1's harmonic decomposition follows from the Gegenbauer expansion of e^{λ t·x} and the eigenfunction relation, and the coefficients c_k,p(λ) are computed from known Bessel-function integrals. The null limit (Theorems 3.1–3.2) is a Hilbert-space CLT plus Karhunen–Loève expansion: the eigenvalues c_k,p(λ)γ_k,p arise from the covariance operator, not from data. The fixed-alternative limit (Theorems 3.3–3.5) uses the SLLN in Hilbert space and Parseval's identity; τ is expressed through the density's harmonic coefficients β_r,k, which are inputs describing the alternative, not fitted parameters. The tuning parameter λ is chosen by the user; the data-driven selection in Section 5.2 is an explicit power-optimization step (oracle or cross-validated) and does not enter the theoretical null/alternative distributions. Self-citations (García-Portugués et al., 2023, 2026; Ebner et al., 2025; Borodavka and Ebner, 2026; Fernández-de-Marcos and García-Portugués, 2023) are used for context, comparison, or standard constructions, never as the proof of the characterization or asymptotics. One non-circular correctness caveat: Proposition 1.1 is stated for any absolutely continuous random vector, but its proof assumes q∈L²(S^{p−1}) ('Let q be the density ... and q∈L²'), and Remark 3.1 then claims consistency 'against all absolutely continuous non-uniform distributions'; the L² restriction is unsupported for densities outside L². This is a domain-of-validity/overclaim issue, not circularity, so the circularity score remains 0.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The central statistical theory is derived from standard harmonic analysis and Hilbert-space limit theorems; the only introduced object is the test function class and the tuning parameter λ. No new physical entities are posited.

free parameters (1)
  • λ (tuning/concentration parameter) = grid search over {i/10: i=1,...,300} / cross-validation
    Tunes the test to a given alternative by maximizing the standardized mean shift; not fitted to the null distribution. In the paper's main power comparisons, the alternative distribution is used to choose λ (oracle), which inflates power.
assumptions (6)
  • standard math Gegenbauer/Chebyshev polynomials form an orthogonal basis of L²([-1,1], (1-u²)^{(p-3)/2} du) and spherical harmonics form a basis of L²(S^{p-1}) (Dai & Xu 2013).
    Used throughout the harmonic decomposition, Funk–Hecke formula, and Karhunen–Loève expansion.
  • standard math Funk–Hecke formula and Bessel integral identities (Zwillinger et al. 2014 Formula 7.321; DLMF 18.3.1, 10.9.2) are used to derive the coefficients m_{k,p}(λ).
    These identities yield the closed-form coefficients c_{k,p}(λ) in Lemma 2.1.
  • standard math Hilbert-space CLT and strong law of large numbers (Henze 2024) are applied to the L²-valued random elements.
    Justifies Theorems 3.1, 3.3, and 3.4.
  • standard math The exponential series e^{λt·x} = Σ m_{k,p}(λ) C_k^{(p-2)/2}(t·x) converges uniformly on [-1,1] (Kalf 1995).
    Allows term-wise application of the Laplace–Beltrami operator and interchange of sums and integrals.
  • domain assumption The test functions {e^{λt·x}: t∈S^{p-1}} for fixed λ>0 form a characterizing class for the uniform law via the Stein operator Δ_{S^{p-1}}.
    Proposition 1.1 proves this, but the sufficiency is a structural property of the exponential class that the entire test relies on.
  • domain assumption Alternative distributions are absolutely continuous with density q ∈ L²(S^{p-1}) w.r.t. uniform measure.
    Needed for the spherical-harmonic expansion (10) to hold in L², which underpins the characterization and the fixed-alternative asymptotic theory.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On Stein's test of uniformity on the hypersphere." pith.science (2026). https://pith.science/paper/3ZLFHRSY

@misc{pith2026260220896,
  author       = {Pith},
  title        = {Pith review of: On Stein's test of uniformity on the hypersphere},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3ZLFHRSY}},
  note         = {Machine review of arXiv:2602.20896}
}
read the original abstract

We propose a new test of uniformity on the hypersphere based on a Stein characterization associated with the Laplace-Beltrami operator. We identify a sufficient class of test functions for this characterization, linked to the moment generating function. Exploiting the operator's eigenfunctions to obtain a harmonic decomposition in terms of Gegenbauer polynomials, we show that the proposed procedure belongs to the class of Sobolev tests. We derive closed-form series representations for the asymptotic distribution of the test statistic under the null hypothesis and under fixed alternatives. To enhance power against a range of alternatives, we introduce a tuning parameter into the characterization and study its impact on rejection probabilities. We discuss data-driven strategies for selecting this parameter to maximize rejection rates for a given alternative and compare the resulting performance with that of related parametric tests. Additional numerical experiments compare the proposed test with competing Sobolev-class procedures, highlighting settings in which it offers clear advantages.

Figures

Figures reproduced from arXiv: 2602.20896 by the authors.

Figure 1
Figure 1. Relative coefficients k 7→ ck,p(λ) and k 7→ c dKSD k,p (λ) for dimensions p = 3 and p = 10, for the L 2 - Stein test (solid lines) and the dKSD test (dashed lines). For each choice of λ, the coefficients are standardized by their maximum. To illustrate the effect, we plot the continuous mappings in k, while in the test, only evaluations at integer values of k are used. To visualize the previous functions, we use the… view at source ↗
Figure 2
Figure 2. Hammer projection representation of s 7→ √ n|z(s)|, for the fixed alternative vMF(µ, κ) and n = 100. The central point is µ = (0, −1, 0)⊤. In the first row, κ = 1 is fixed, while in the second, λ = 1 is. (a) λ = 0.1 (b) λ = 1 (c) λ = 10 [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Hammer projection representation of the null correlation kernel s 7→ ρ(s, t), for t = (0, 0, 1)⊤ (north pole, diamond). The shape of the kernel is invariant from the choice of t. 5.2 Parameter selection The testing procedure can be adapted to a specific alternative by selecting the tuning parameter λ that maximizes the rejection rate under the underlying distribution. To optimize the parameter λ we consider the stan… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Hammer projection representation of the fixed-alternative correlation kernel s 7→ ρ ′ (s, t), for the fixed alternative vMF(µ, κ) and t = (0, 0, 1)⊤ (north pole, diamond). The central point is µ = (0, −1, 0)⊤. The concentration parameter λ only affects the test statist…
Figure 5
Figure 5. Figure 5: Powers of Stein, softmax tests, and dKSD test, with concentration parameter λ in each column under the alternative distributions MvMF(10), SCM(3), and W(2). In the top row, the simulation was carried out in dimension p = 3 while the bottom row was carried out in dimens…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 3 linked inside Pith

  1. [1]

    E., Ghaderinezhad, F., Gorham, J., Gretton, A., Ley, C., Liu, Q., Mackey, L., Oates, C

    Anastasiou, A., Barp, A., Briol, F.-X., Ebner, B., Gaunt, R. E., Ghaderinezhad, F., Gorham, J., Gretton, A., Ley, C., Liu, Q., Mackey, L., Oates, C. J., Reinert, G., and Swan, Y. (2023). Stein's method meets computational statistics: A review of some recent developments. Stat. Sci. , 38(1):120--139

    2023

  2. [2]

    Barbour, A. D. (1988). Stein's method and P oisson process convergence. J. Appl. Probab. , 25(A):175--184

    1988

  3. [3]

    Barbour, A. D. (1990). Stein's method for diffusion approximations. Probab. Theory Relat. Fields , 84(3):297--322

    1990

  4. [4]

    Baringhaus, L., Ebner, B., and Henze, N. (2017). The limit distribution of weighted L ^2 -goodness-of-fit statistics under fixed alternatives, with applications. Ann. Inst. Stat. Math. , 69(5):969--995

    2017

  5. [5]

    J., Porcu, E., and Girolami, M

    Barp, A., Oates, C. J., Porcu, E., and Girolami, M. (2022). A Riemann--Stein kernel method. Bernoulli , 28(4):2181--2208

    2022

  6. [6]

    Beran, R. J. (1968). Testing for uniformity on a compact homogeneous space. J. Appl. Probab. , 5(1):177--195

    1968

  7. [7]

    Bingham, C. (1974). An antipodally symmetric distribution on the sphere. Ann. Stat. , 2(6):1201--1225

    1974

  8. [8]

    Borodavka, J. I. and Ebner, B. (2026). A general maximal projection approach to uniformity testing on the hypersphere . Bernoulli , 32(2):996 -- 1019

    2026

Show all 38 references
  1. [9]

    Cai, T., Fan, J., and Jiang, T. (2013). Distributions of angles in random packing on spheres. J. Mach. Learn. Res. , 14(21):1837--1864

    2013

  2. [10]

    Chen, L. H. Y., Goldstein, L., and Shao, Q.-M. (2011). Normal Approximation by S tein's Method . Springer, Berlin, Heidelberg

    2011

  3. [11]

    Cutting, C., Paindaveine, D., and Verdebout, T. (2017). Testing uniformity on high-dimensional spheres against monotone rotationally symmetric alternatives. Ann. Stat. , 45(3):1024--1058

    2017

  4. [12]

    and Xu, Y

    Dai, F. and Xu, Y. (2013). Approximation Theory and Harmonic Analysis on Spheres and Balls . Springer Monographs in Mathematics. Springer, New York

    2013

  5. [13]

    Ding, Y., Markatou, M., and Saraceno, G. (2025). Poisson kernel-based tests for uniformity on the \(d\) -dimensional sphere. Stat. Sin. , 35(4):1947--1969

    2025

  6. [14]

    NIST Digital Library of Mathematical Functions

    DLMF (2020). NIST Digital Library of Mathematical Functions . http://dlmf.nist.gov/, Release 1.0.27 of 2020-06-15. F. W. J. Olver, A. B. Olde Daalhuis , D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller and B. V. Saunders, eds

    2020

  7. [15]

    Ebner, B., García-Portugués, E., and Verdebout, T. (2025). High-dimensional S obolev tests on hyperspheres. arXiv:2501.10898

    2025 arXiv

  8. [16]

    and Garc\'ia-Portugu\'es, E

    Fern\'andez-de-Marcos , A. and Garc\'ia-Portugu\'es, E. (2023). On new omnibus tests of uniformity on the hypersphere. Test , 32(4):1508--1529

    2023

  9. [17]

    E., and Swan, Y

    Fischer, A., Gaunt, R. E., and Swan, Y. (2026). Stein’s method of moments on the sphere . Bernoulli , 32(2):1186--1212

    2026

  10. [18]

    Garc\'ia-Portugu\'es, E., Navarro-Esteban, P., and Cuesta-Albertos, J. A. (2023). On a projection-based class of uniformity tests on the hypersphere. Bernoulli , 29(1):181--204

    2023

  11. [19]

    Garc\'ia-Portugu\'es, E., Paindaveine, D., and Verdebout, T. (2026). On a class of S obolev tests for symmetry, their detection thresholds, and asymptotic powers. To appear: J. Am. Stat. Assoc

    2026

  12. [20]

    and Verdebout, T

    Garc\'ia-Portugu\'es, E. and Verdebout, T. (2018). A review of uniformity tests on the hypersphere. arXiv:1804.00286

    2018 arXiv

  13. [21]

    Gin\'e, E. (1975). Invariant tests for uniformity on compact R iemannian manifolds based on S obolev norms. Ann. Stat. , 3(6):1243--1266

    1975

  14. [22]

    Gregory, G. G. (1977). Large sample theory for U -statistics and tests of fit. Ann. Stat. , 5(1):110--123

    1977

  15. [23]

    Henze, N. (2024). Asymptotic Stochastics : An Introduction with a View towards Statistics , volume 10 of Mathematics Study Resources . Springer, Berlin, Heidelberg

    2024

  16. [24]

    Hsu, E. P. (2002). Stochastic Analysis on Manifolds , volume 38 of Graduate Studies in Mathematics . American Mathematical Society

    2002

  17. [25]

    Kalf, H. (1995). On the expansion of a function in terms of spherical harmonics in arbitrary dimensions. Bull. Belg. Math. Soc. Simon Stevin , 2(4):361--380

    1995

  18. [26]

    and Verdebout, T

    Ley, C. and Verdebout, T. (2017). Modern Directional Statistics . Chapman & Hall/CRC Interdisciplinary Statistics Series. CRC Press, Boca Raton

    2017

  19. [27]

    Linnik, Y. V. (1953a). Linear forms and statistical criteria. I . Ukrainskii Matematicheskii Zhurnal , 5:207--243. In Russian

  20. [28]

    Linnik, Y. V. (1953b). Linear forms and statistical criteria. II . Ukrainskii Matematicheskii Zhurnal , 5:247--290. In Russian

  21. [29]

    Mardia, K. V. and Jupp, P. E. (1999). Directional Statistics . Wiley Series in Probability and Statistics. Wiley, Chichester

    1999

  22. [30]

    Nikitin, Y. Y. (2017). Tests based on characterizations, and their efficiencies: A survey. Acta Comment. Univ. Tartu. Math. , 21(1):34--55

    2017

  23. [31]

    and Garc\'ia-Portugu\'es, E

    Pewsey, A. and Garc\'ia-Portugu\'es, E. (2021). Recent advances in directional statistics. Test , 30(1):1--58

    2021

  24. [32]

    Prentice, M. J. (1978). On invariant tests of uniformity for directions and orientations. Ann. Stat. , 6(1):169--176

    1978

  25. [33]

    and Vemuri, B

    Qu, X. and Vemuri, B. C. (2025). Theory and applications of kernel S tein's method on R iemannian manifolds. arXiv:2501.00695

    2025 arXiv

  26. [34]

    Rayleigh, Lord . (1919). On the problem of random vibrations, and of random flights in one, two, or three dimensions. Lond. Edinb. Dublin Philos. Mag. J. Sci. , 37(220):321--347

    1919

  27. [35]

    Stein, C. M. (1972). A bound for the error in the normal approximation to the distribution of a sum of dependent random variables. In Proceedings of the S ixth B erkeley S ymposium on M athematical S tatistics and P robability ( U niv. C alifornia, B erkeley, C alif., 1970/197...

    1972

  28. [36]

    and Matsuda, T

    Xu, W. and Matsuda, T. (2020). A S tein goodness-of-fit test for directional distributions. In Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics , volume 108 of Proceedings of Machine Learning Research , pages 320--330. PMLR

    2020

  29. [37]

    and Matsuda, T

    Xu, W. and Matsuda, T. (2021). Interpretable S tein goodness-of-fit tests on R iemannian manifold. In Proceedings of the 38th International Conference on Machine Learning , volume 139 of Proceedings of Machine Learning Research , pages 11502--11513. PMLR

    2021

  30. [38]

    Table of Integrals, Series, and Products

    Zwillinger, D., Moll, V., Gradshteyn, I., and Ryzhik, I., editors (2014). Table of Integrals, Series, and Products . Academic Press, Boston, eighth edition

    2014

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.