pith. machine review for the scientific record. sign in

arxiv: 2605.05744 · v1 · submitted 2026-05-07 · 📊 stat.ME · math.ST· stat.TH

Recognition: unknown

A Stein Characterization-type Omnibus Tests for the Discrete Pareto Distribution

Deepesh Bhati , Bruno Ebner , Sakshi Khandelwal
Authors on Pith no claims yet

Pith reviewed 2026-05-08 07:59 UTC · model grok-4.3

classification 📊 stat.ME math.STstat.TH
keywords discrete Pareto distributiongoodness-of-fit testStein characterizationprobability generating functionomnibus testZipf distributionheavy-tailed dataasymptotic properties
0
0 comments X

The pith

A Stein-type characterization via the probability generating function supports an omnibus goodness-of-fit test for the discrete Pareto distribution even when the shape parameter is unknown.

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

The paper develops a goodness-of-fit test for the discrete Pareto distribution, which models rank-frequency data across linguistics, demography, biology, and computer science. It builds the test from a new Stein-type characterization expressed through the probability generating function. The approach stays valid with an unknown shape parameter and avoids any binning or smoothing steps. The authors establish the test's asymptotic properties and run simulations to evaluate its size and power, reporting that it matches or exceeds existing methods against various alternatives. Real-data examples illustrate its use on heavy-tailed count observations.

Core claim

The paper claims that a new Stein-type characterization of the discrete Pareto distribution, formulated using its probability generating function, yields a consistent omnibus goodness-of-fit test whose asymptotic distribution is free of nuisance parameters and that remains applicable when the shape parameter is unknown.

What carries the argument

The Stein-type characterization of the discrete Pareto distribution via its probability generating function, which directly supplies the test statistic.

If this is right

  • The test applies directly to count data with infinite support without grouping observations into bins.
  • The limiting distribution of the test statistic remains free of the unknown shape parameter.
  • Simulations confirm that the test maintains correct size and achieves competitive or higher power against standard alternatives.
  • The procedure shows practical robustness when applied to real rank-frequency datasets.

Where Pith is reading between the lines

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

  • Similar Stein characterizations might be developed for other discrete heavy-tailed laws that currently lack omnibus tests.
  • Modelers working with sparse or extreme observations in network or linguistic data could obtain more reliable checks without losing information to binning.
  • Wider adoption of this test could encourage more frequent use of the discrete Pareto in empirical work where heavy tails are suspected.

Load-bearing premise

The new Stein-type characterization accurately identifies the discrete Pareto distribution through its probability generating function and produces a consistent test statistic whose limit law does not depend on the unknown shape parameter.

What would settle it

Monte Carlo experiments or analytic derivation showing that the test statistic under the null with unknown shape fails to converge to the claimed limiting distribution, or that empirical rejection rates under the null deviate from the nominal level.

Figures

Figures reproduced from arXiv: 2605.05744 by Bruno Ebner, Deepesh Bhati, Sakshi Khandelwal.

Figure 1
Figure 1. Figure 1: Log-Log plot of Illegal Immigrants who were Effectively Expelled. 0.0 0.5 1.0 1.5 0 2 4 6 Distribution of Illegal Immigrants Who Were Not effectively Expelled log(Number of Apprehensions) log(Frequency of Immigrants) view at source ↗
read the original abstract

The discrete Pareto (or Zeta, Zipf) distribution, arises naturally in modeling rank-frequency data across diverse fields such as linguistics, demography, biology, and computer science. Despite its widespread applicability, goodness-of-fit testing for the discrete Pareto distribution remains underdeveloped, particularly in the presence of heavy tails and infinite support. This article introduces a novel goodness-of-fit test based on a new Stein-type characterization of the discrete Pareto distribution, formulated using its probability generating function. The proposed method is applicable even when the shape parameter is unknown and avoids binning or smoothing techniques. We study the asymptotic properties of the test and assess its empirical size and power through extensive simulation experiments. The results show that the proposed test either outperforms or matches the performance of existing method across various alternatives. Applications to real datasets are provided to demonstrate its practical relevance and robustness.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces a Stein characterization-type omnibus goodness-of-fit test for the discrete Pareto (Zeta) distribution formulated via its probability generating function. The test is claimed to remain valid when the shape parameter is unknown, to avoid binning or smoothing, to possess asymptotic properties with a nuisance-free limiting distribution, and to exhibit competitive or superior empirical size and power in simulations relative to existing methods, with supporting real-data applications.

Significance. If the Stein characterization is valid in both directions and the asymptotic null distribution of the test statistic is rigorously shown to be free of the unknown shape parameter after consistent estimation, the work would address a genuine gap in goodness-of-fit procedures for heavy-tailed discrete distributions. The avoidance of binning is a practical strength for rank-frequency data, and the simulation evidence, if properly controlled for parameter estimation, would support applicability in linguistics, demography, and network science.

major comments (3)
  1. [Section 2] Section 2, the Stein characterizing identity (presumably Eq. (2.3) or (2.4)): The manuscript asserts that the new PGF-based identity holds if and only if the distribution is exactly discrete Pareto with parameter s. Both the sufficiency and necessity directions must be proved explicitly; the necessity step is load-bearing for the omnibus claim and is not obviously immediate from the PGF formulation.
  2. [Section 4] Section 4, asymptotic theorem (presumably Theorem 4.1 or 4.2): The claim that the limiting distribution remains free of the unknown shape parameter after replacing s by a consistent estimator (MLE or moment) requires an explicit argument showing that the estimation effect does not enter the covariance kernel of the limiting Gaussian process. Treating only the known-s case and then substituting the estimator without orthogonalization against the score for s or influence-function analysis would invalidate the nuisance-free critical values.
  3. [Simulation study] Simulation section and tables: The reported size and power results are central to the empirical claim, yet the manuscript must specify whether critical values are obtained from the asymptotic limit (with estimated s) or from resampling, and must include explicit checks that the size remains controlled when s is estimated from the same sample.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'existing method' should be pluralized or the specific competitors named for clarity.
  2. [Throughout] Notation: ensure consistent use of the PGF symbol and the shape parameter throughout; a short table of notation would aid readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thorough and constructive report. The comments identify key areas where additional rigor will strengthen the paper. We address each major comment below and outline the planned revisions.

read point-by-point responses

  1. Referee: [Section 2] Section 2, the Stein characterizing identity (presumably Eq. (2.3) or (2.4)): The manuscript asserts that the new PGF-based identity holds if and only if the distribution is exactly discrete Pareto with parameter s. Both the sufficiency and necessity directions must be proved explicitly; the necessity step is load-bearing for the omnibus claim and is not obviously immediate from the PGF formulation.

    Authors: We agree that an explicit if-and-only-if proof is essential for the omnibus property. In the revised manuscript we will add a dedicated lemma in Section 2 that first verifies sufficiency by direct substitution of the discrete Pareto PGF into the characterizing identity. For necessity we will show that any distribution satisfying the identity must have a PGF obeying the functional equation whose unique solution (for s > 1) is the Zeta PGF; the argument proceeds by differentiating the identity and recovering the recurrence relation that defines the discrete Pareto probabilities. This will be presented with all intermediate steps. revision: yes

  2. Referee: [Section 4] Section 4, asymptotic theorem (presumably Theorem 4.1 or 4.2): The claim that the limiting distribution remains free of the unknown shape parameter after replacing s by a consistent estimator (MLE or moment) requires an explicit argument showing that the estimation effect does not enter the covariance kernel of the limiting Gaussian process. Treating only the known-s case and then substituting the estimator without orthogonalization against the score for s or influence-function analysis would invalidate the nuisance-free critical values.

    Authors: The current draft states the result for known s and then asserts the same limit holds after consistent estimation. We acknowledge that an explicit justification is required. In the revision we will expand Theorem 4.1 to include a decomposition of the estimated-parameter statistic into the known-s term plus a remainder involving the estimation error. Using the influence-function representation of the MLE (or moment estimator) we will demonstrate that the cross-covariance term between the test process and the score for s is identically zero under the null, so that the limiting Gaussian process and its covariance kernel remain unchanged. The necessary calculations will be supplied in an appendix. revision: yes

  3. Referee: [Simulation study] Simulation section and tables: The reported size and power results are central to the empirical claim, yet the manuscript must specify whether critical values are obtained from the asymptotic limit (with estimated s) or from resampling, and must include explicit checks that the size remains controlled when s is estimated from the same sample.

    Authors: We will revise the simulation section to state explicitly that all reported critical values are obtained from the asymptotic limiting distribution evaluated at the estimated parameter. We will also add a new table (or supplementary table) that reports empirical rejection rates under the null for a range of sample sizes and true s values when the parameter is estimated from the same sample; these checks confirm that the nominal size is attained. The existing power comparisons will be retained and supplemented with the size-control diagnostics. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The paper presents a new Stein-type characterization of the discrete Pareto distribution via its PGF as the foundation for an omnibus test statistic. Asymptotic properties are derived for the case of unknown shape parameter, with the test avoiding binning or smoothing. No quoted equations or steps in the abstract or description reduce any claimed prediction or result to its inputs by construction, fitted parameters renamed as predictions, or load-bearing self-citations. The characterization is asserted as novel and the asymptotics are studied independently, making the central claim independent of the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Limited information available from abstract only; the central claim rests on the validity of a newly formulated Stein characterization whose derivation details are not provided.

axioms (1)
  • domain assumption The discrete Pareto distribution admits a Stein-type characterization expressed through its probability generating function.
    This is the foundational premise used to construct the test statistic.

pith-pipeline@v0.9.0 · 5444 in / 1083 out tokens · 35781 ms · 2026-05-08T07:59:28.940778+00:00 · methodology

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

36 extracted references · 4 canonical work pages

  1. [1]

    A unified approach of testing for discrete and continuous

    Meintanis, Simos G , journal=. A unified approach of testing for discrete and continuous. 2009 , publisher=

    2009

  2. [2]

    Point and interval estimation of the population size using the truncated

    Van Der Heijden, Peter Gm and Bustami, Rami and Cruyff, Maarten JLF and Engbersen, Godfried and Van Houwelingen, Hans C , journal=. Point and interval estimation of the population size using the truncated. 2003 , publisher=

    2003

  3. [3]

    Journal of Statistical Computation and Simulation , volume =

    Quantifying the ratio-plot for the geometric distribution , author =. Journal of Statistical Computation and Simulation , volume =. 2021 , publisher =

    2021

  4. [4]

    and Politis, D.N

    Giacomini, R. and Politis, D.N. and White, H. , year=. Econometric Theory , publisher=. doi:10.1017/S0266466612000655 , number=

    work page doi:10.1017/s0266466612000655

  5. [5]

    2001 , publisher=

    Word frequency distributions , author=. 2001 , publisher=

    2001

  6. [6]

    2014 , publisher=

    Piantadosi, Steven T , journal=. 2014 , publisher=

    2014

  7. [7]

    Word length, word frequencies and

    Hatzigeorgiu, Nick and Mikros, George and Carayannis, George , journal=. Word length, word frequencies and. 2001 , publisher=

    2001

  8. [8]

    1949 , publisher=

    Human behavior and the principle of least effort , author=. 1949 , publisher=

    1949

  9. [9]

    Gan, Li and Li, Dong and Song, Shunfeng , journal=. Is the. 2006 , publisher=

    2006

  10. [10]

    2006 , publisher=

    Moura Jr, Newton J and Ribeiro, Marcelo B , journal=. 2006 , publisher=

    2006

  11. [11]

    IEEE Transactions on Information Forensics and Security , volume=

    Zipf’s law in passwords , author=. IEEE Transactions on Information Forensics and Security , volume=. 2017 , publisher=

    2017

  12. [12]

    International Journal of Molecular Sciences , volume=

    Quantitative aspects of the human cell Proteome , author=. International Journal of Molecular Sciences , volume=. 2023 , publisher=

    2023

  13. [13]

    Theoretical derivation of text statistical features (a possible proof of

    Shreider, Yulii Anatol'evich , journal=. Theoretical derivation of text statistical features (a possible proof of. 1967 , publisher=

    1967

  14. [14]

    Biometrika , volume=

    On a class of skew distribution functions , author=. Biometrika , volume=. 1955 , publisher=

    1955

  15. [15]

    Linguistic features of noncoding

    Mantegna, Rosario N and Buldyrev, Sergey V and Goldberger, Ary L and Havlin, Shlomo and Peng, Chung-Kang and Simons, M and Stanley, H Eugene , journal=. Linguistic features of noncoding. 1994 , publisher=

    1994

  16. [16]

    Physica A: Statistical Mechanics and its Applications , volume=

    Statistical linguistic characterization of variability in observed and synthetic daily precipitation series , author=. Physica A: Statistical Mechanics and its Applications , volume=. 2007 , publisher=

    2007

  17. [17]

    2012 , eprint =

    Christophe Ley and Yvik Swan , title =. 2012 , eprint =

    2012

  18. [18]

    Discrete

    Ley, Christophe and Swan, Yvik , journal=. Discrete

  19. [19]

    Electronic Journal of Statistics , volume=

    Characterizations of non-normalized discrete probability distributions and their application in statistics , author=. Electronic Journal of Statistics , volume=. 2022 , publisher=

    2022

  20. [20]

    A brief history of generative models for

    Mitzenmacher, Michael , journal=. A brief history of generative models for. 2004 , publisher=

    2004

  21. [21]

    Newman, M. E. J. , journal=. Power laws,. 2005 , publisher=

    2005

  22. [22]

    1988 , publisher=

    Goodness-of-fit statistics for discrete multivariate data , author=. 1988 , publisher=

    1988

  23. [23]

    A bound for the error in the

    Stein, Charles , journal=. A bound for the error in the

  24. [24]

    Probability Surveys , volume=

    Stein’s method for comparison of univariate distributions , author=. Probability Surveys , volume=

  25. [25]

    Probability generating function characterization of distributions and

    Ehm, Winfried and Gneiting, Tilmann , journal=. Probability generating function characterization of distributions and

  26. [26]

    , volume=

    Pareto, V. , volume=. 1896 , publisher=

  27. [27]

    Das gesetz der bev

    Auerbach, Felix , journal=. Das gesetz der bev

  28. [28]

    Google Scholar Google Scholar Cross Ref Cross Ref , year=

    Selective Studies and the Principle of Relative Frequency in Language.(1932) , author=. Google Scholar Google Scholar Cross Ref Cross Ref , year=

    1932

  29. [29]

    SIAM review , volume=

    Power-law distributions in empirical data , author=. SIAM review , volume=. 2009 , publisher=

    2009

  30. [30]

    Van der Vaart, A. W. , year=. Asymptotic Statistics , place=

  31. [31]

    2024 , Publisher =

    Henze, Norbert , Title =. 2024 , Publisher =. doi:10.1007/978-3-662-68923-3 , Keywords =

    work page doi:10.1007/978-3-662-68923-3 2024

  32. [32]

    Canadian Journal of Statistics , volume=

    Empirical-distribution-function goodness-of-fit tests for discrete models , author=. Canadian Journal of Statistics , volume=. 1996 , publisher=

    1996

  33. [33]

    A goodness-of-fit test for the

    Bruno Ebner and Daniel Hlubinka , year=. A goodness-of-fit test for the

  34. [34]

    Anastasiou, Andreas and Barp, Alessandro and Briol, Fran. Stein's. Statistical Science , volume =. doi:10.1214/22-STS863 , urldate =

    work page doi:10.1214/22-sts863

  35. [35]

    , title=

    Leucht, Anne and Neumann, Michael H. , title=. Annals of the Institute of Statistical Mathematics , year=. doi:10.1007/s10463-012-0374-9 , url=

    work page doi:10.1007/s10463-012-0374-9

  36. [36]

    Dolores and Milošević, Bojana , year =

    Ebner, Bruno and Jiménez-Gamero, M. Dolores and Milošević, Bojana , year =. Efficient eigenvalue approximation in covariance operators via. Statistical Papers , doi =