pith. sign in

arxiv: 2606.19306 · v1 · pith:K3LJHS5Ynew · submitted 2026-06-17 · 🧮 math.PR · math.ST· stat.TH

On two overlooked stick-breaking constructions of the normalized inverse Gaussian process

Annalisa Cerquetti This is my paper

Pith reviewed 2026-06-26 19:38 UTC · model grok-4.3

classification 🧮 math.PR math.STstat.TH
keywords normalized inverse Gaussian processstick-breaking constructionBayesian nonparametricBrownian excursiongeneralized Gamma subordinatorPoisson-Kingman modelsPoisson-Gamma processes
0
0 comments X

The pith

Two stick-breaking constructions for the normalized inverse Gaussian process follow from earlier results on Brownian excursions and generalized Gamma subordinators.

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

The paper identifies two previously overlooked stick-breaking representations for the normalized inverse Gaussian random discrete distribution. One arises by mixing the conditional Brownian excursion partition over the local time at zero up to time one. The other is obtained as a special case of a construction for priors based on random spatial and temporal change of the normalized generalized Gamma subordinator. Both use straightforward transformations of standard random variables and extend to wider families of mixed Poisson-Kingman models and Poisson-Gamma processes.

Core claim

The normalized inverse Gaussian process admits two alternative stick-breaking constructions: one derived from the Aldous-Pitman result on conditional Brownian excursion partitions by mixing over local time at zero up to time one, and the other as a particular case of James' result for priors from random spatial and temporal change of the normalized generalized Gamma subordinator. Both are straightforward transformations of standard random variables.

What carries the argument

Stick-breaking representations obtained by mixing over local time in conditional Brownian excursion partitions and by spatial-temporal changes of normalized generalized Gamma subordinators.

If this is right

  • The constructions supply explicit simulation methods for the NIG process via transformations of ordinary random variables.
  • The first construction extends to any mixed Poisson-Kingman model driven by the 1/2 stable Levy measure.
  • The second construction extends to any Poisson-Gamma process driven by the inverse Gaussian subordinator.
  • Alternative sampling routes become available for these families in Bayesian nonparametric models.

Where Pith is reading between the lines

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

  • Cross-referencing results from excursion theory with Bayesian nonparametric literature may produce further representations for other normalized Levy processes.
  • The explicit forms could simplify comparison of computational efficiency across different stick-breaking schemes for related random discrete distributions.
  • Generalizations might yield new parameter-free sampling algorithms for processes driven by stable or inverse Gaussian subordinators.

Load-bearing premise

The cited results from Aldous and Pitman (1998) and James (2013) translate directly into valid stick-breaking representations for the NIG process in the Bayesian nonparametric setting without further conditions.

What would settle it

Explicit computation of the marginal distributions of the weights produced by the two proposed transformations, showing they differ from those of the normalized inverse Gaussian process.

read the original abstract

We shed light on two alternative stick-breaking constructions of the normalized inverse Gaussian (NIG) random discrete distribution which appear to have been overlooked so far in the Bayesian nonparametric setting. The first is derived from a result in Aldous and Pitman (1998) for the conditional Brownian excursion partition, mixing over the local time at zero up to time one. The second arises as a particular case of a result in James (2013) for priors obtained by a random spatial and temporal change of the normalized generalized Gamma subordinator. Both constructions are in terms of straightforward transformations of standard random variables and can be easily generalized to provide the stick-breaking construction of any element, respectively, in a) the family of mixed Poisson-Kingman models driven by the $1/2$ stable L\'evy measure and b) the family of Poisson-Gamma processes driven by the Inverse Gaussian subordinator.

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

2 major / 2 minor

Summary. The manuscript identifies two stick-breaking constructions for the normalized inverse Gaussian (NIG) random discrete distribution that have been overlooked in the Bayesian nonparametric literature. The first is obtained by mixing the conditional Brownian excursion partition from Aldous and Pitman (1998) over the local time at zero up to time one. The second is recovered as a special case of the random spatial and temporal change of the normalized generalized Gamma subordinator in James (2013). The paper further indicates that the same approach yields stick-breaking representations for the broader family of mixed Poisson-Kingman models driven by the 1/2-stable Lévy measure and for Poisson-Gamma processes driven by the inverse-Gaussian subordinator.

Significance. If the claimed derivations are valid, the constructions supply explicit representations of the NIG weights via transformations of standard random variables, which may simplify simulation and theoretical analysis within the Poisson-Kingman and related classes. The explicit links to excursion theory and subordinator transformations also clarify connections between existing results that had not been previously exploited for the NIG case.

major comments (2)
  1. [Abstract / Introduction] The central claim that the Aldous-Pitman (1998) conditional Brownian excursion partition, after mixing over local time at zero up to time 1, yields a stick-breaking representation of the NIG process, requires explicit verification that the resulting weights match the known NIG marginals or the 1/2-stable Poisson-Kingman family. The abstract states the construction but does not display the mixing measure or the resulting Lévy measure.
  2. [Abstract / Section on James (2013) construction] Similarly, the claim that the NIG arises as a particular case of James (2013) requires the manuscript to specify the exact parameter settings (spatial and temporal change functions) that recover the inverse-Gaussian subordinator after normalization, and to confirm that no additional unstated conditions are needed.
minor comments (2)
  1. [Title] The title refers to the 'normalized inverse Gaussian process' while the abstract and body emphasize the 'random discrete distribution'; a brief clarifying sentence on the distinction would improve precision.
  2. [Abstract] The generalization statements to the 1/2-stable Poisson-Kingman and Poisson-Gamma families are stated at the end of the abstract; a short dedicated paragraph or remark indicating the precise parameter restrictions would make the scope clearer.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and for identifying points where additional explicitness would strengthen the presentation. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract / Introduction] The central claim that the Aldous-Pitman (1998) conditional Brownian excursion partition, after mixing over local time at zero up to time 1, yields a stick-breaking representation of the NIG process, requires explicit verification that the resulting weights match the known NIG marginals or the 1/2-stable Poisson-Kingman family. The abstract states the construction but does not display the mixing measure or the resulting Lévy measure.

    Authors: Section 2 of the manuscript derives the mixing measure over local time at zero up to time 1 and verifies that the resulting weights match the 1/2-stable Poisson-Kingman family by direct comparison of the Lévy measure. The abstract is kept concise as is conventional, but we will revise the introduction to display the mixing measure and the recovered Lévy measure explicitly. revision: yes

  2. Referee: [Abstract / Section on James (2013) construction] Similarly, the claim that the NIG arises as a particular case of James (2013) requires the manuscript to specify the exact parameter settings (spatial and temporal change functions) that recover the inverse-Gaussian subordinator after normalization, and to confirm that no additional unstated conditions are needed.

    Authors: Section 3 specifies the exact spatial and temporal change functions from James (2013) that recover the inverse-Gaussian subordinator (with the normalization step yielding the NIG) and states that these are direct special cases without further conditions. We will add these parameter settings to the abstract and introduction for clarity. revision: yes

Circularity Check

0 steps flagged

No circularity: constructions explicitly derived from independent external results

full rationale

The paper states that the two stick-breaking constructions are 'derived from a result in Aldous and Pitman (1998)' and 'arise as a particular case of a result in James (2013)'. These are external citations to prior independent work by different authors. No self-citations are load-bearing, no parameters are fitted within the paper and then renamed as predictions, and no equations reduce by construction to inputs defined in the present manuscript. The central claim concerns translation of external results into the NIG setting, which is an issue of correctness or completeness rather than circularity. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper relies on standard results from probability theory on Brownian excursions, Levy processes, and subordinators without introducing new free parameters, axioms beyond those in the citations, or invented entities.

axioms (2)
  • standard math Results from Aldous and Pitman (1998) on conditional Brownian excursion partitions and local time mixing
    Invoked directly to derive the first construction.
  • standard math Results from James (2013) on random spatial and temporal changes of normalized generalized Gamma subordinators
    Invoked directly to derive the second construction.

pith-pipeline@v0.9.1-grok · 5677 in / 1420 out tokens · 30257 ms · 2026-06-26T19:38:22.528254+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

17 extracted references · 1 canonical work pages

  1. [1]

    Pitman, J

    Aldous, D. Pitman, J. (1998) The standard additive coalescent.Ann. Probab.26 (4) 1703-1726

    1998

  2. [2]

    Canale, A., Corradin, R., Nipoti, B. (2022). Importance conditional sampling for Pit- man Yor mixtures.Statistics and Computing, 32(3) 1-18

    2022

  3. [3]

    Walker, S.G

    Favaro, S. Walker, S.G. (2013) Slice Samplingσ-Stable Poisson-Kingman Mixture Mod- els.JCGS, 22, 4, 830-847 9

    2013

  4. [4]

    (2012) On the Stick-Breaking Representation of Normalized Inverse Gaussian Priors.Biometrika, 99, 663-674

    Favaro, S., Lijoi, A., and Pr¨ unster, I. (2012) On the Stick-Breaking Representation of Normalized Inverse Gaussian Priors.Biometrika, 99, 663-674

    2012

  5. [5]

    and Teh, Y

    Favaro, S., Lomeli, M., Nipoti, B. and Teh, Y. W. (2014) On the Stick-breaking rep- resentation ofσ-stable Poisson-Kingman models.Electron. J. Statist.8 (1) 1063 - 1085

    2014

  6. [6]

    Lijoi, A

    Favaro, S. Lijoi, A. Nava, C. Nipoti, B. Pr¨ unster, Teh, Y.I.W. (2016) On the Stick- Breaking Representation for Homogeneous NRMIs.Bayesian Analysis11 (3) 697 - 724

    2016

  7. [7]

    (1973) A Bayesian Analysis of Some Nonparametric Problems.Ann

    Ferguson, T.S. (1973) A Bayesian Analysis of Some Nonparametric Problems.Ann. Statist.1 (2) 209 - 230. https://doi.org/10.1214/aos/1176342360

    work page doi:10.1214/aos/1176342360 1973

  8. [8]

    (2005), Exchangeable Gibbs Partitions and Stirling Trian- gles, Zapiski Nauchnych Seminarov POMI, 325, 83-102

    Gnedin, A., and Pitman, J. (2005), Exchangeable Gibbs Partitions and Stirling Trian- gles, Zapiski Nauchnych Seminarov POMI, 325, 83-102. [834]

    2005

  9. [9]

    James, L. F. (2013) Stick-breaking PG(α, ζ)-Generalized Gamma Processes.arXiv:1308.6570 [math-PR]

    Pith/arXiv arXiv 2013

  10. [10]

    James, L. F. (2019) Stick-breaking Pitman-Yor processes given the species sampling size.arXiv:1908.07186 [math.ST]

    arXiv 2019

  11. [11]

    H., and Pr¨ unster, I

    Lijoi, A., Mena, R. H., and Pr¨ unster, I. (2005) Hierarchical Mixture Modelling With Normalized Inverse-Gaussian Priors.JASA, 100, 1278-1291

    2005

  12. [12]

    (2003) Self-similar fragmentations and stable sub- ordinators

    Miermont, G., and Schweinsberg, J. (2003) Self-similar fragmentations and stable sub- ordinators. In S ˜A¨minaire de Probabilit ˜A¨ XXXVII, vol. 1832 of LNh in Maths. Springer, Berlin, pp. 333-359

    2003

  13. [13]

    and Yor, M

    Perman, M., Pitman, J. and Yor, M. (1992) Size-biased sampling of Poisson point processes and excursions.Probab. Th. Rel. Fields92, 21-39

    1992

  14. [14]

    (1995) Exchangeable and partially exchangeable random partitions.Probab

    Pitman, J. (1995) Exchangeable and partially exchangeable random partitions.Probab. Th. Rel. Fields102, 145-158

    1995

  15. [15]

    (2003) Poisson-Kingman Partitions, in Science and Statistics: A Festschrift for Terry Speed, ed

    Pitman, J. (2003) Poisson-Kingman Partitions, in Science and Statistics: A Festschrift for Terry Speed, ed. D. R. Goldstein, Beachwood, OH: IMS, pp. 1-34

    2003

  16. [16]

    Pitman and M. Yor. (1992) Arcsine laws and interval partitions derived from a stable subordinator. Proceedings of the London Mathematical Society (3), 65:326-356, 1992

    1992

  17. [17]

    (1997), The Two Parameter Poisson-Dirichlet Distribution Derived From a Stable Subordinator.Annals of Probability, 27, 1870-1902 10

    Pitman, J., and Yor, M. (1997), The Two Parameter Poisson-Dirichlet Distribution Derived From a Stable Subordinator.Annals of Probability, 27, 1870-1902 10

    1997