Pith. sign in

REVIEW 2 major objections 5 minor 40 references

Uncentered counts of large blocks in Pitman–Yor partitions converge to an explicit mixture of infinitely divisible laws.

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

2026-07-30 23:09 UTC pith:VWXZW4CX

load-bearing objection Solid Gibbs-to-Pitman-Yor spectrum limits with a real but fixable gap when they pass from conditional local limits to the marginal in Theorem 3.2. the 2 major comments →

arxiv 2607.23401 v1 pith:VWXZW4CX submitted 2026-07-26 math.PR math.STstat.MEstat.TH

Limit Theorems for the Pitman-Yor Frequency Spectrum

Ross Arthur Maller , Soudabeh Shemehsavar This is my paper
classification math.PR math.STstat.MEstat.TH MSC 60C0560F0560G5105A18
keywords random partitionscomponent frequency spectrumPitman-Yor sampling formulaGibbs distributionlimit shapesallele frequency spectruminfinitely divisible laws
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 studies how the frequency spectrum of a random partition of {1,…,n} behaves for large n under the two-parameter Pitman–Yor model. It first derives a general moment-generating-function identity that expresses linear combinations of the spectrum, jointly with the number of blocks, in terms of triangular arrays of i.i.d. random variables. Specialising to Pitman–Yor, the authors prove that the uncentered number of blocks whose sizes lie between λ n and μ n converges in distribution to an explicit mixture involving a stable-type subordinator truncated to (0,1] and a sum of i.i.d. size-biased jumps. The same machinery recovers the classical local limit theorem for the total number of blocks and yields conditional limits given that number. The results sit in the “convergent” regime of partition-shape theory, where ordinary limit shapes fail to exist, and they suggest a possible functional limit theorem for the cumulative spectrum process.

Core claim

For the Pitman–Yor sampling formula the un-normalised partial sums S_λ,n = ∑_{j=⌊λ n⌋}^n M_jn converge in distribution to an explicit proper probability mass function that is a mixture, over a Mittag-Leffler-type weight, of the density at 1 of a truncated stable subordinator plus an independent sum of ℓ i.i.d. size-biased jumps on [λ,1]. The same identity produces joint and conditional limits for general linear functionals of the spectrum.

What carries the argument

The mgf identity of Theorem 2.1: the joint transform of Mn/Kn and Kn is written as the product of an ordinary mgf of centred i.i.d. V-variables and a ratio of local probabilities for two triangular arrays of i.i.d. ÛX-variables. All subsequent Pitman–Yor limits are obtained by verifying Kallenberg conditions for those arrays and justifying the local-limit passage to densities at the point 1.

Load-bearing premise

The passage from characteristic-function convergence of the triangular-array sums to pointwise convergence of their densities at the single point 1 must hold uniformly enough to interchange limit and integral; if that local-limit step fails for some admissible weight functions the density ratios that appear in every main theorem become unjustified.

What would settle it

Numerically sample large-n Pitman–Yor partitions for fixed (α,θ,λ) and compare the empirical distribution of S_λ,n against the explicit integral formula of Theorem 3.2; systematic discrepancy for moderate ℓ would falsify the claimed limit.

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

If this is right

  • The allele-frequency spectrum functionals used in genetics (homozygosity, site-frequency spectrum bins) possess explicit large-sample distributions under Pitman–Yor sampling without further centring or scaling.
  • Finite-dimensional distributions of the cumulative process λ ↦ S_λ,n are available by the same mgf identity, opening a route to a functional limit theorem.
  • The same triangular-array representation applies, with only notational changes, to other Gibbs-type and Poisson–Kingman partitions once their weight sequences q_j are known.
  • In the convergent regime of partition-shape theory one obtains genuine distributional limits rather than the Gaussian fluctuations characteristic of the expansive regime.

Where Pith is reading between the lines

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

  • The appearance of the same truncated-stable characteristic exponent that governs ratios of trimmed subordinators suggests a deeper link between the Pitman–Yor spectrum and the jump structure of stable processes that could be made rigorous by Poisson-point-process methods.
  • Because the limiting pmf does not depend on the second Pitman–Yor parameter θ after conditioning on the number of blocks, many genetic summary statistics may be asymptotically ancillary for θ.
  • The general identity of Theorem 2.1 supplies a practical Monte-Carlo scheme: simulate the i.i.d. arrays rather than the full partition, which may be cheaper for very large n.

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 / 5 minor

Summary. The paper studies the component frequency spectrum (M_{jn}) of Gibbs-type random partitions, with detailed analysis of the two-parameter Pitman–Yor sampling model. Theorem 2.1 gives an exact mgf identity for u_n^T(M_n/K_n − q_n) jointly with {K_n=k}, expressing it as a centered mgf of i.i.d. V-variables times a ratio of point probabilities of sums of triangular arrays of X̂-variables, times P(K_n=k). For the Pitman–Yor model, Theorem 3.1 establishes a joint local limit theorem with an explicit limit involving the density at 1 of an infinitely divisible law; Corollaries 3.1–3.2 specialize to power weights (recovering the Ewens–Watterson statistic at p=2) and to counts S^{λ,µ}_n = Σ_{j=⌊λn⌋}^{⌊µn⌋} M_{jn} given K_n = ⌊xn^α⌋. Theorem 3.2, the second main result, gives the marginal limit distribution of S^λ_n as an explicit proper pmf — a mixture of x-integrated densities of Y_x(α,λ)+H_ℓ(λ) at 1. Proofs use a Poissonization/multinomial identity (Lemma 5.1), Kallenberg's array conditions, a Gnedenko–Kolmogorov-style local limit argument with a uniform exponential cf bound, and mgf asymptotics under the integrability condition (3.3).

Significance. If correct, the paper provides new, explicit, parameter-free limit laws for the uncentered, unnormed frequency spectrum of the Pitman–Yor partition — a regime distinct from the Gaussian "expansive-case" limits of Erlihson–Granovsky [5] (the authors themselves locate their results in the "convergent" category and explain the contrast). Theorem 2.1 is a clean general identity for Gibbs partitions of independent interest, and the derivation contains genuine internal checks: u_n=1_n and p=1 reduce to known identities (§5), and (3.6) recovers Pitman's local limit theorem for K_n, which the paper re-proves by a new route. The explicit limiting pmf in Theorem 3.2 is a falsifiable, simulable expression with direct relevance to allele-frequency-spectrum statistics in genetics. The connection noted to Kevei–Mason [24] in Remark (4.68)–(4.69) is an interesting bonus.

major comments (2)
  1. [§4, proof of Theorem 3.2] Proof of Theorem 3.2, (4.58)–(4.62): the passage from the fixed-x joint local limit (4.59)/(4.60) to the marginal pmf (3.10) integrates over 0<x<∞ without justifying the interchange of the n-limit and the x-integral. Pointwise convergence for each fixed x does not control mass escaping toward x=0, x=∞, or moving x-ranges; Lemma 4.3's Fatou step covers K_n alone and yields only a one-sided bound, not domination for the joint (ℓ,x) density. This step converts the conditional Corollary 3.2 into the marginal Theorem 3.2, so it is load-bearing. A repair appears available within the paper's own tools: p_n(ℓ,x) = n^α P(S^λ_n=ℓ, K_n=⌊xn^α⌋) are densities on {0,1,...}×(0,∞) (counting×Lebesgue) summing to 1, and the limit is shown proper in §5; a Scheffé-type argument would then give L¹ and hence marginal convergence. Note the properness proof currently derives (4.66) via the same unproven integra
  2. [§4, proof of Corollary 3.2] Proof of Corollary 3.2, (4.49)–(4.53): Feller's Laplace inversion operator (4.51) is applied, and limit, τ-integral, and the inversion/ℓ-summation are interchanged to obtain (4.52)–(4.53), with the only justification the remark that summing (4.53) over ℓ gives 1 via (4.54) 'and the interchange is valid'. Since (3.9) feeds both Corollary 3.2 and Theorem 3.2, this needs an actual argument. Bounds of the type (5.12)/(4.63) for the full exponent including g(x,iτ,λ,µ) (e.g., |e^{-g}g^ℓ| ≤ C_ℓ uniformly, with |E(e^{iτY^{(0)}_x})| ≤ e^{-cx|τ|^α}) should make the τ-integrations and interchanges routine; please supply the details.
minor comments (5)
  1. [§3, Theorem 3.1] The hypotheses on the weight function f (bounded variation, or continuity a.e., together with (3.3) and f(0)=0) are stated in §1 and used in the proof of Lemma 4.1, but Theorem 3.1's statement does not list them. Please restate the full assumptions on f in Theorem 3.1 so the theorem is self-contained.
  2. [General] Several typographical slips: double periods in the abstract and after (3.1); (4.16) conditions on 'K_n = n' (should be K_n = k); (4.7) defines Â^{(J)}_{kn} with '1 < j ≤ J' (should be 1 ≤ j ≤ J); (4.57) has an unmatched parenthesis and should read (xc(λ,µ))^ℓ E(...); in the proof of Theorem 3.1, 'f_{Y(ν,f)_θ}(1)' should be f_{Y(ν,f)_x}(1); 'Lebesque' → 'Lebesgue'.
  3. [§5, derivation of Theorem 3 of [26]] In (5.29), the second line P(V^{(u_n)}_{1n} = 0) = q_{jn} for J+1 ≤ j ≤ n should presumably be P(V^{(u_n)}_{1n}=0) = Σ_{j>J} q_{jn}; as written each j>J is assigned probability q_{jn} for the same value 0. Please clarify.
  4. [References] Reference [2] lacks volume/page information; [11] is cited only as an arXiv preprint — update if published. In [33] 'characterizedby' is missing a space.
  5. [§3–§4] It would help the reader to add one sentence after Corollaries 3.1–3.2 explaining why θ-independence of the conditional limits is to be expected (conditioning on K_n = ⌊xn^α⌋ fixes the θ-dependent factor), and to state explicitly in Theorem 3.2 that the lattice endpoints in the x-discretization contribute negligibly to (4.58).

Circularity Check

0 steps flagged

No circularity: limits are derived from the classical Pitman–Yor formula via triangular-array local limits, not forced by definition or self-citation.

full rationale

The paper starts from the standard Gibbs form (1.2) and the classical two-parameter Pitman–Yor sampling formula (3.1), derives a general joint mgf identity (Theorem 2.1) by Poissonization/multinomial representation (Lemma 5.1), and obtains the new limit laws (Theorems 3.1–3.2, Corollaries 3.1–3.2) by verifying Kallenberg conditions and a Gnedenko–Kolmogorov local-limit argument for the triangular arrays of ˆX and V. Self-citations ([26], [11], etc.) supply background or are re-derived as checks (Appendix recovery of Theorem 3 of [26]; identification of the Mittag-Leffler density); they do not define the target marginals or force the Lévy measures. There are no fitted parameters renamed as predictions, no uniqueness theorems imported to forbid alternatives, and no ansatz smuggled in via citation. The skeptic’s concern about interchanging limit and x-integral in the proof of Theorem 3.2 is a possible analytic gap, not circularity. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 6 axioms · 0 invented entities

Load-bearing inputs are standard: the Pitman–Yor / Gibbs sampling formula, classical triangular-array and Fourier local-limit tools, and an integrability condition on the weight function f. No parameters are fitted to data. No new physical or combinatorial entities are postulated beyond ordinary infinitely divisible laws built from truncated stable Lévy measures.

axioms (6)
  • domain assumption The partition law has Gibbs form (1.2): P(Mn=m,Kn=k)=C_nk ∏ (q_j^{m_j}/m_j!) on the composition simplex A_kn.
    Starting point of Theorem 2.1; specialized to Pitman–Yor q_j in (3.2).
  • domain assumption Pitman–Yor two-parameter sampling formula (3.1) with 0<α<1, θ>−α.
    Section 3; classical model from Pitman–Yor 1997, taken as given.
  • ad hoc to paper Weight functions f satisfy ∫_0^1 y^{−α−1}|f(y)| dy < ∞ and are of bounded variation (or continuous a.e.) on [0,1] with f(0)=0.
    Condition (3.3) and the dominated-convergence argument in Lemma 4.1; needed for V-array mgf limits and integral approximations.
  • standard math Kallenberg’s criteria for convergence of row-i.i.d. triangular arrays to infinitely divisible limits (Cor. 15.16 of Kallenberg 2002).
    Used to identify Y^{(ν,f)}_x in Lemma 4.1.
  • standard math Fourier inversion and lattice local-limit tail bounds in the style of Gnedenko–Kolmogorov.
    Display (4.17) and the three-region split (5.7)–(5.8) justifying density convergence at 1.
  • domain assumption Kn(α,θ)/n^α → Mittag-Leffler a.s. / in distribution (Pitman).
    Used to choose conditioning levels k_n=⌊x n^α⌋ and to identify g_α in (4.37)–(4.38).

pith-pipeline@v1.2.0-grok45-kimik3 · 32990 in / 3151 out tokens · 61932 ms · 2026-07-30T23:09:28.955332+00:00 · methodology

0 comments
read the original abstract

We derive a general distribution formula applicable to a wide variety of Gibbs-type partitions and use it to obtain large sample results for linear combinations of the component frequency spectrum $(M_{jn})_{1\le j\le n}$ (in genetics, the allele frequency spectrum) associated with a random partitioning of $\{1,2,\ldots, n\}$. The two-parameter Pitman-Yor sampling model is analysed in detail and asymptotic distributions of sums of the form $\sum _{j=\lf \lambda n\rf}^{\lf \mu n\rf} M_{jn}$, $0<\lambda\le \mu\le 1$, are obtained. Our results suggest a possible functional limit theorem for $\sum _{j=\lf \lambda n\rf}^{n} M_{jn}$. Useful connections with limit shapes for random structures on the set of partitions and other applications are suggested.

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

40 extracted references · 2 canonical work pages

  1. [1]

    & Tavaré, S

    Arratia, R., Barbour, A.D. & Tavaré, S. (2003)Logarithmic Combinatorial Structures: A Probabilistic Approach.EMS Mono. Math., Euro. Math. Soc., Zurich

    2003

  2. [2]

    & Zarepour, M

    Chegini, S. & Zarepour, M. (2023) Random discrete probability measures based on a negative binomial process.Canadian Journal of Statistics,

    2023

  3. [3]

    (1952) The influence of the maximum term in the addition of independent random variables.Trans

    Darling, D.A. (1952) The influence of the maximum term in the addition of independent random variables.Trans. Amer. Math. Soc.,73, 95–107

    1952

  4. [4]

    & Gorodetsky, O

    Elboim, D. & Gorodetsky, O. (2024) Multiplicative arithmetic functions and the generalized Ewens measure.Israel J. Mathematics 262, 143–189

    2024

  5. [5]

    & Granovsky, B.L

    Erlihson, M.M. & Granovsky, B.L. (2008) Limit shapes of Gibbs distributions on the set of integer partitions: the expansive case.Annales de l’Inst. Henri Poincaré Probab. Statist. 44, 915–945

    2008

  6. [6]

    (1972) The sampling theory of selectively neutral alleles.Theoret

    Ewens, W. (1972) The sampling theory of selectively neutral alleles.Theoret. Pop. Biol., 3, 87–112

    1972

  7. [7]

    (1979)Mathematical Population Genetics, Springer-Verlag, NY

    Ewens, W. (1979)Mathematical Population Genetics, Springer-Verlag, NY

    1979

  8. [8]

    (1971)An Introduction to Probability Theory and Its Applications, II.Wiley, New York

    Feller, W. (1971)An Introduction to Probability Theory and Its Applications, II.Wiley, New York

    1971

  9. [9]

    & Kolmogorov, A.N

    Gnedenko, B.V. & Kolmogorov, A.N. (1968)Limit Distributions for Sums of Independent Random Variables.Addison-Wesley

    1968

  10. [10]

    & Pitman, J

    Gnedin, A., Hansen, B. & Pitman, J. (2007) Notes on the occupancy problem with infinitely many boxes: general asymptotics and power laws.Probability Surveys, 4, 146–171

    2007

  11. [11]

    & Shemehsavar, S

    Griffiths, R.C., Maller, R.A. & Shemehsavar, S. (2024) A Gibbs sampling scheme for a generalised Poisson-Kingman classArxiv 2402.11563

    Pith/arXiv arXiv 2024

  12. [12]

    & Lau, J.W

    Ho, M.W., James, L.F. & Lau, J.W. (2021) Gibbs partitions, Riemann-Liouville fractional operators, Mittag-Leffler functions, and fragmentations derived from stable subordinators.J. Appl. Prob., 58, 314–334

    2021

  13. [13]

    & Lau, J.W

    Ho, M.W., James, L.F. & Lau, J.W. (2025) Inverse clustering of Gibbs partitions via independent fragmentationanddualdependentcoagulationoperatorsJ. Appl. Prob., 1-18. doi:10.1017/jpr.2025.28

    work page doi:10.1017/jpr.2025.28 2025

  14. [14]

    & Tucker, H.G

    Hudson, W.N. & Tucker, H.G. (1975). On admissible translates of infinitely divisible distributions.Z. Wahrscheinlichkeitstheorie verw. Gebiete 32, 65- 72

    1975

  15. [15]

    & Maller, R.A

    Ipsen, Y.F., Kevei, P. & Maller, R.A. (2018) Convergence to stable limits for ratios of trimmed Levy processes and their jumps.Markov Processes and Related Fields, 24, 539–562

    2018

  16. [16]

    & Maller, R.A

    Ipsen, Y.F. & Maller, R.A. (2017) Negative binomial construction of random discrete distributions on the infinite simplex.Theor. Stoch. Proc., 22, 34–46. 22

    2017

  17. [17]

    & Shemehsavar (2020a) Limiting distributions of generalised Poisson- Dirichlet distributions based on negative binomial processes.J

    Ipsen, Y.F., Maller, R.A. & Shemehsavar (2020a) Limiting distributions of generalised Poisson- Dirichlet distributions based on negative binomial processes.J. Theor. Prob., 33, 1974–2000

    1974

  18. [18]

    & Shemehsavar, S

    Ipsen, Y.F., Maller, R.A. & Shemehsavar, S. (2020b) Size biased sampling from the Dickman sub- ordinator.Stoch. Proc. Appl., 130, 6880–6900

  19. [19]

    & Shemehsavar, S

    Ipsen, Y.F., Maller, R.A. & Shemehsavar, S. (2021) A generalised Dickman distribution and the number of species in a negative binomial process model.Adv. Appl. Prob., 53, 370–399

    2021

  20. [20]

    (2008) Large sample asymptotics for the two-parameter Poisson-Dirichlet process In: Pushing the Limits of Contemporary Statistics: Contributions in Honor of Jayanta K

    James, L.F. (2008) Large sample asymptotics for the two-parameter Poisson-Dirichlet process In: Pushing the Limits of Contemporary Statistics: Contributions in Honor of Jayanta K. GhoshIMS Collections, Vol. 3 (2008) 187–199

    2008

  21. [21]

    (2025) Poisson hierarchical Indian buffet processes – with indications for microbiome species sampling models.arXiv: 2502.01919v1 [stat.ML]

    James, L.F. (2025) Poisson hierarchical Indian buffet processes – with indications for microbiome species sampling models.arXiv: 2502.01919v1 [stat.ML]

    Pith/arXiv arXiv 2025

  22. [22]

    (2002)Foundations of Modern Probability, 2nd Ed.Springer

    Kallenberg, O. (2002)Foundations of Modern Probability, 2nd Ed.Springer

    2002

  23. [23]

    Kesten, Harry (1963) Ratio theorems for random walks II.J. Anal. Math.11, 323–379

    1963

  24. [24]

    & Mason, D.M

    Kevei, P. & Mason, D.M. (2014) The limit distribution of ratios of jumps and sums of jumps of subordinators.ALEA, Latin Amer. J. Probab. Math. Stat., 11, 631–642

    2014

  25. [25]

    & Shemehsavar, S

    Maller, R.A. & Shemehsavar, S. (2023) Generalized Poisson–Dirichlet distributions based on the Dickman subordinator.Theor. Prob. Appl., 67, 593–612

    2023

  26. [26]

    & Shemehsavar, S

    Maller, R.A. & Shemehsavar, S. (2025) Asymptotics of the allele frequency spectrum and the number of alleles.J. Appl. Prob., 62, 516–540. doi:10.1017/jpr.2024.84

    work page doi:10.1017/jpr.2024.84 2025

  27. [27]

    (2017) Extreme sizes in Gibbs-type exchangeable random partitions.Ann

    Mano, S. (2017) Extreme sizes in Gibbs-type exchangeable random partitions.Ann. Inst. Stat. Math., 69, 1–37

    2017

  28. [28]

    (2006a) On sampling distributions for coalescent processes with simultaneous multiple collisions.Bernoulli, 12, 35–53

    Möhle, M. (2006a) On sampling distributions for coalescent processes with simultaneous multiple collisions.Bernoulli, 12, 35–53

  29. [29]

    (2006b) On the number of segregating sites for populations with large family sizes

    Möhle, M. (2006b) On the number of segregating sites for populations with large family sizes. Adv. Appl. Prob. 38, 750–767

  30. [30]

    & Zeindler, D

    Nikeghbali, A. & Zeindler, D. (2013) The generalized weighted probability measure on the symmetric group and the asymptotic behavior of the cycles.Annales de l’Inst. Henri Poincaré Probab. Statist. 49, 961–981

    2013

  31. [31]

    & Yor, M

    Perman, M., Pitman, J. & Yor, M. (1992) Size-biased sampling of Poisson point processes and excursions.Probab. Theor. Rel. Fields, 92, 21–39

    1992

  32. [32]

    (1997) Partition structures derived from Brownian motion and stable subordinators

    Pitman, J. (1997) Partition structures derived from Brownian motion and stable subordinators. Bernoulli, 3, 79–96

    1997

  33. [33]

    (1999) Brownian motion, bridge, excursion, and meander characterizedby sampling at independent uniform times.Electron

    Pitman, J. (1999) Brownian motion, bridge, excursion, and meander characterizedby sampling at independent uniform times.Electron. J. Probab.4, 1–33

    1999

  34. [34]

    (2006)Combinatorial Stochastic Processes

    Pitman, J. (2006)Combinatorial Stochastic Processes. Springer-Verlag, Berlin

    2006

  35. [35]

    & Yor, M

    Pitman, J. & Yor, M. (1997) The two-parameter Poisson–Dirichlet distribution derived from a stable subordinator.Ann. Probab., 25, 855–900

    1997

  36. [36]

    Sato, K. I. (1999)Lévy Processes and Infinitely Divisible Distributions. Cambridge University Press, Cambridge

    1999

  37. [37]

    (2024) Gibbs partitions: a comprehensive phase diagram.Annales de l’Institut Henri Poincaré Probabilités et Statistiques, 60, 1729–1766

    Stufler, B. (2024) Gibbs partitions: a comprehensive phase diagram.Annales de l’Institut Henri Poincaré Probabilités et Statistiques, 60, 1729–1766. 23

    2024

  38. [38]

    & Sagae, M

    Tanabe, K. & Sagae, M. (1992) An exact Cholesky decomposition and the generalized inverse of the variance-covariance matrix of the multinomial distribution, with applications.J. Roy. Statist. Soc. B (Method.), 54, 211–219

    1992

  39. [39]

    (1977) Heterosis or neutrality?Genetics, 85, 789–814

    Watterson, G.A. (1977) Heterosis or neutrality?Genetics, 85, 789–814

    1977

  40. [40]

    (2000) Moments of some statistics of Pitman sampling formula.Bull

    Yamato, H., Sibuya, M. (2000) Moments of some statistics of Pitman sampling formula.Bull. Inf. Cybern. 32, 1–10. R.A. Maller, Research School of Finance, Actuarial Studies & Statistics The Australian National University, Canberra, ACT, 0200, Australia. Email: Ross.Maller@anu.edu.au S. Shemehsavar, College of Science, Technology, Engineering & Mathematics ...

    2000