pith. sign in

arxiv: 2604.11363 · v1 · submitted 2026-04-13 · 🧮 math.ST · stat.TH

Subordinated Wright-Fisher Priors

Nathan A. Judd , Dario Span\`o This is my paper

Pith reviewed 2026-05-10 16:26 UTC · model grok-4.3

classification 🧮 math.ST stat.TH
keywords subordinated Wright-FisherDirichlet priorsstochastic time-changedual processespartially exchangeable datatime-dependent priorsdiffusion processesBayesian nonparametrics
0
0 comments X

The pith

A random time-change of the Wright-Fisher diffusion yields a new class of Dirichlet priors that allow jumps and non-Markovian memory while preserving exact conjugacy.

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

The paper constructs time-dependent Dirichlet priors by subordinating the standard Wright-Fisher diffusion with an independent stochastic clock that acts as a hyper-prior on operational time. This produces priors on evolving allele frequencies whose paths can jump and retain memory of earlier states. For partially exchangeable observations recorded at discrete times, the authors supply explicit representations of the prior and posterior together with exact simulation algorithms for both the frequency process and the driving clock. The constructions rest on a new family of discrete dual processes that restore conjugacy after the time change. A reader would care because many genetic or compositional data sets exhibit abrupt shifts or path dependence that continuous Markov diffusions cannot accommodate without losing tractability.

Core claim

By subordinating a Wright-Fisher diffusion with an independent stochastic time-change interpreted as a hyper-prior on the operational clock, one obtains a class of time-dependent Dirichlet priors that accommodate discontinuous trajectories and non-Markovian memory. Explicit representations and exact sampling algorithms are derived for the prior and posterior distributions of both the process and the clock when the data are partially exchangeable and observed at discrete time points. Computability and conjugacy are achieved through a novel class of discrete dual processes that generalise earlier duality results for Wright-Fisher models.

What carries the argument

Stochastic subordination that replaces deterministic time with a random clock in the Wright-Fisher diffusion, made tractable by a new family of discrete dual processes that preserve conjugacy under partial exchangeability.

If this is right

  • Exact sampling algorithms become available for both the subordinated process and its driving clock under discrete-time partially exchangeable data.
  • Trajectories may contain discontinuities while all finite-dimensional distributions remain Dirichlet.
  • Non-Markovian memory effects can be introduced without sacrificing posterior conjugacy or computability.
  • The framework extends classical Wright-Fisher conjugacy results to a broader class of time-inhomogeneous priors.

Where Pith is reading between the lines

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

  • Such priors could be applied to model sudden population shifts or allele-frequency jumps induced by environmental events.
  • The discrete dual processes may generalise to other neutral diffusions such as the Moran model.
  • One could examine whether inferred clocks on real genetic time series reveal interpretable patterns of memory or jump clustering.

Load-bearing premise

The chosen stochastic time-change together with the new discrete dual processes must preserve the Dirichlet marginals and conjugacy structure for partially exchangeable discrete-time observations.

What would settle it

A dataset of partially exchangeable discrete-time observations for which the posterior of the frequency process and clock cannot be sampled exactly from the dual-process representations, or whose finite-dimensional distributions deviate from the claimed Dirichlet form.

Figures

Figures reproduced from arXiv: 2604.11363 by Dario Span\`o, Nathan A. Judd.

Figure 1
Figure 1. Figure 1: Laplace transform maps t 7→ E[e −λnC(t) ] for α-stable (α = 0.5), Gamma process (b = 1), and tempered-stable (α = 0.7, λ = 0.5), subordi￾nators, and their inverse subordinator counterparts, computed for, clockwise from the topleft: (θ, n) = (1, 2),(1, 5),(0.2, 2),(0.2, 5). The solid lines show the analytical Laplace transforms of the subordinators, the dashed lines the numerically-computed Laplace transfor… view at source ↗
Figure 2
Figure 2. Figure 2: 100 Simulated trajectories of the sWF’s dual process for [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: 100, 000 samples from the sWF transition function pe (θ,C) t (0.5, dz) with θ = (0.5, 0.5), t ∈ {0.25, 0.5, 1, 2} and four subordinator clocks C = S, (clockwise from topleft): (a) the identity S(t) ≡ t, (b) α-Stable subordinator with α = 0.5; (c) Inverse Gaussian with δ = γ = 1; (d) Gamma process with a = b = 1 19 [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Smoothed density estimate [using Kernel Density Estimation] (left) [PITH_FULL_IMAGE:figures/full_fig_p020_4.png] view at source ↗
Figure 1
Figure 1. Figure 1: Laplace transform maps t → E[e −λnC(t) ] across the range t ∈ [0, 2] and for a selection of (n, θ) and subordinators: (a) α-Stable with α = 0.5;(b) Poisson c = 1; (c) Inverse Gaussian, δ = γ = 1; (d) Gamma a = b = 1; (e) identity. 1 [PITH_FULL_IMAGE:figures/full_fig_p029_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: 100, 000 samples from the sWF transition function pe (θ,C) t (x, dz) with mutation parameters θ = (1, 1), initial coordinate x = 1/2 and time-input t ∈ {0.25, 0.5, 1, 2} where each pane represent a different subordinator clock C = S, (clockwise from topleft): (a) the identity S(t) ≡ t, (b) α-Stable subordinator with α = 0.5; (c) Inverse Gaussian with δ = γ = 1; (d) Gamma process with a = b = 1. 2 [PITH_FU… view at source ↗
read the original abstract

A new class of time-dependent Dirichlet priors is introduced as a generalisation of the Wright-Fisher diffusion, allowing discontinuities in the trajectories, as well as non-Markovian memory. This class is obtained as a simple stochastic time-change (subordination), interpreted as a hyper-prior assigned to the operational time-clock of a Wright-Fisher diffusion. Explicit representations and exact sampling algorithms are obtained for prior and posterior distributions of the process and of its clock, given partially exchangeable data sampled at discrete time-points. Computability and conjugacy rely on a novel class of discrete dual processes, generalising existing results on duality and computable filters.

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

0 major / 2 minor

Summary. The paper introduces a new class of time-dependent Dirichlet priors obtained by subordinating the Wright-Fisher diffusion via a stochastic time-change, interpreted as a hyper-prior on the operational clock. This construction yields processes with discontinuous trajectories and non-Markovian memory. Explicit representations and exact sampling algorithms are claimed for the prior and posterior distributions of both the process and the clock, given partially exchangeable data observed at discrete time points. Conjugacy and computability are achieved through a novel class of discrete dual processes that generalize existing duality results for diffusions.

Significance. If the central claims on explicit representations and exact algorithms hold, the work would meaningfully extend duality-based filtering techniques to subordinated, non-Markovian processes. This could provide tractable Bayesian nonparametric priors for time-varying population models or exchangeable data with temporal dependence, building directly on known conjugacy properties of the Wright-Fisher diffusion while adding flexibility through the random time change.

minor comments (2)
  1. The abstract asserts 'explicit representations' and 'exact sampling algorithms' without indicating where in the manuscript the derivations or pseudocode appear; adding forward references to the relevant sections or theorems would improve readability.
  2. Notation for the subordinating process and the dual discrete process should be introduced with a clear table or diagram early in the paper to distinguish the operational time from the calendar time.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of the manuscript, including the accurate summary of the contributions and the encouraging evaluation of its significance. We appreciate the recommendation for minor revision and will incorporate improvements to enhance clarity, presentation, and any minor issues in the next version.

Circularity Check

0 steps flagged

Derivation self-contained; no circular reductions identified

full rationale

The paper constructs a subordinated Wright-Fisher process via stochastic time-change and obtains explicit prior/posterior representations through a novel discrete dual that generalizes known duality techniques. No equation equates a derived quantity to its input by construction, no fitted parameter is relabeled as a prediction, and no load-bearing step reduces to an unverified self-citation chain. The central claims of conjugacy and computability rest on the mathematical properties of the subordination and dual rather than tautological redefinition or renaming of known results.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Insufficient detail in abstract to identify specific free parameters, axioms, or invented entities; subordination and dual processes are presented as the core technical contributions but their foundational assumptions are not enumerated.

pith-pipeline@v0.9.0 · 5391 in / 1190 out tokens · 43046 ms · 2026-05-10T16:26:10.941923+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

45 extracted references · 45 canonical work pages

  1. [1]

    Antoniak

    Charles E. Antoniak. Mixtures of Dirichlet Processes with Applications to Bayesian Nonparametric Problems.The Annals of Statistics, 2(6):1152 – 1174, 1974

    work page 1974

  2. [2]

    Levy processes and stochastic calculus.Cambridge Studies in Advanced Mathematics, 116, 2009

    D Applebaum. Levy processes and stochastic calculus.Cambridge Studies in Advanced Mathematics, 116, 2009

    work page 2009

  3. [3]

    Cede˜ no-Gir´ on, and Bruno Toaldo

    Ivan Bioˇ ci´ c, Daniel E. Cede˜ no-Gir´ on, and Bruno Toaldo. Sampling inverse subordinators and subdiffusions, 2025

    work page 2025

  4. [4]

    S. Bochner. Diffusion equation and stochastic processes.Proceedings of the National Academy of Sciences, 35(7):368–370, 1949. 25

    work page 1949

  5. [5]

    N. Bouleau. Quelques resultats probabilistes sur la subordination au sens de bochner. In Francis Hirsch and Gabriel Mokobodzki, editors,S´ eminaire de Th´ eorie du Potentiel Paris, No. 7, pages 54–81, Berlin, Heidelberg, 1984. Springer Berlin Heidelberg

    work page 1984

  6. [6]

    Inference in hidden Markov models

    Olivier Capp´ e, ´Eric Moulines, and Tobias Ryd´ en. Inference in hidden Markov models. InSpringer series in statistics, 2005

    work page 2005

  7. [7]

    Amin, Alex Ali, Joshua Rollins, Sebastian W

    Nuria Alina Chandra, Yucen Lily Li, Alan N. Amin, Alex Ali, Joshua Rollins, Sebastian W. Ober, Aniruddh Raghu, and Andrew Gordon Wilson. A unification of discrete, gaussian, and simplicial diffusion, 2025

    work page 2025

  8. [8]

    Exact Simulation of Truncated L´ evy Subordinator.ACM Transactions on Modeling and Computer Simulation, 30, 10 2019

    Angelos Dassios, Jia Lim, and Yan Qu. Exact Simulation of Truncated L´ evy Subordinator.ACM Transactions on Modeling and Computer Simulation, 30, 10 2019

    work page 2019

  9. [9]

    New York: Springer-Verlag, 1986

    Luc Devroye.Non-Uniform Random Variate Generation. New York: Springer-Verlag, 1986

    work page 1986

  10. [10]

    The Ja- cobi diffusion process as a neuronal model.Chaos, 28, 10 2018

    Giuseppe D’Onofrio, Massimiliano Tamborrino, and Petr Lansky. The Ja- cobi diffusion process as a neuronal model.Chaos, 28, 10 2018

    work page 2018

  11. [11]

    Etheridge.Some Mathematical Models from Population Genetics, vol- ume 2012

    A. Etheridge.Some Mathematical Models from Population Genetics, vol- ume 2012. Springer, 2011

    work page 2012

  12. [12]

    Etheridge and R

    A. Etheridge and R. Griffiths. A coalescent dual process in a Moran model with genic selection.Theoretical population biology, 75:320–30, 2009

    work page 2009

  13. [13]

    S. N. Ethier and R. C. Griffiths. The Transition Function of a Fleming-Viot Process.The Annals of Probability, 21(3):1571–1590, 1993

    work page 1993

  14. [14]

    Ewens.Mathematical population genetics / Warren J

    Warren John. Ewens.Mathematical population genetics / Warren J. Ewens.Biomathematics ; Vol.9. Springer-Verlag, Berlin ;, 1979

    work page 1979

  15. [15]

    New York: John Wiley & Sons, 2nd edition, 1971

    William Feller.An Introduction to Probability Theory and its Applications, volume II. New York: John Wiley & Sons, 2nd edition, 1971

    work page 1971

  16. [16]

    A non-linear explicit filter.Statistics & Proba- bility Letters, 61(2):145–154, 2003

    Valentine Genon-Catalot. A non-linear explicit filter.Statistics & Proba- bility Letters, 61(2):145–154, 2003

    work page 2003

  17. [17]

    Fast exact simulation of the first-passage event of a subordinator, 06 2023

    Jorge Ignacio Gonz´ alez C´ azares, Feng Lin, and Aleksandar Mijatovic. Fast exact simulation of the first-passage event of a subordinator, 06 2023

    work page 2023

  18. [18]

    Multivariate jacobi process with application to smooth transitions.Journal of Econometrics, 131(1):475– 505, 2006

    Christian Gourieroux and Joann Jasiak. Multivariate jacobi process with application to smooth transitions.Journal of Econometrics, 131(1):475– 505, 2006

    work page 2006

  19. [19]

    Griffiths

    R. Griffiths. A transition density expansion for a multi-allele diffusion model.Advances in Applied Probability, 11(2):310–325, 1979. 26

    work page 1979

  20. [20]

    Griffiths

    R. Griffiths. Coalescent lineage distributions.Advances in Applied Proba- bility, 38(2):405–429, 2006

    work page 2006

  21. [21]

    Griffiths and D

    R. Griffiths and D. Span` o. Orthogonal polynomial kernels and canonical correlations for Dirichlet measures.Bernoulli, 19(2):548–598, 2013

    work page 2013

  22. [22]

    R. C. Griffiths and D. Span` o.Diffusion processes and coalescent trees, page 358–379. London Mathematical Society Lecture Note Series. Cambridge University Press, 2010

    work page 2010

  23. [23]

    Griffiths and Wen-Hsiung Li

    Robert C. Griffiths and Wen-Hsiung Li. Simulating allele frequencies in a population and the genetic differentiation of populations under mutation pressure.Theoretical Population Biology, 23(1):19–33, 1983

    work page 1983

  24. [24]

    Dual process in the two-parameter poisson-dirichlet diffusion.arXiv preprint arXiv:2102.08520, 2021

    Robert C Griffiths, Matteo Ruggiero, Dario Span` o, and Youzhou Zhou. Dual process in the two-parameter poisson-dirichlet diffusion.arXiv preprint arXiv:2102.08520, 2021

    work page arXiv 2021

  25. [25]

    Hamilton

    James D. Hamilton. Analysis of time series subject to changes in regime. Journal of Econometrics, 45(1):39–70, 1990

    work page 1990

  26. [26]

    Alan G. Hawkes. Spectra of some self-exciting and mutually exciting point processes.Biometrika, 58(1):83–90, 1971

    work page 1971

  27. [27]

    Alan G. Hawkes. Hawkes processes and their applications to finance: a review.Quantitative Finance, 18(2):193–198, 2018

    work page 2018

  28. [28]

    Jenkins and Dario Span` o

    Paul A. Jenkins and Dario Span` o . Exact simulation of the Wright–Fisher diffusion.The Annals of Applied Probability, 27(3), jun 2017

    work page 2017

  29. [29]

    World Scientific Pub., 2012

    Panki Kim, Renming Song, and Zoran Vondraˇ cek.Potential Theory of Subordinate Brownian Motions Revisited, pages 243–290. World Scientific Pub., 2012

    work page 2012

  30. [30]

    Exact inference for a class of hidden Markov models on general state spaces

    Guillaume Kon Kam King, Omiros Papaspiliopoulos, and Matteo Ruggiero. Exact inference for a class of hidden Markov models on general state spaces. Electronic Journal of Statistics, 15(1):2832 – 2875, 2021

    work page 2021

  31. [31]

    V. N. Kolokoltsov. Generalized continuous-time random walks, subordina- tion by hitting times, and fractional dynamics.Theory of Probability & Its Applications, 53(4):594–609, 2009

    work page 2009

  32. [32]

    Leonenko, Mark M

    Nikolai N. Leonenko, Mark M. Meerschaert, and Alla Sikorskii. Fractional Pearson diffusions.Journal of Mathematical Analysis and Applications, 403(2):532–546, 2013

    work page 2013

  33. [33]

    Imperial College Press, 2010

    Francesco Mainardi.Fractional calculus and waves in linear viscoelasticity. Imperial College Press, 2010

    work page 2010

  34. [34]

    Fractional l´ evy processes with an application to long memory moving average processes.Bernoulli, 12(6):1099–1126, 2006

    Tina Marquardt. Fractional l´ evy processes with an application to long memory moving average processes.Bernoulli, 12(6):1099–1126, 2006. 27

    work page 2006

  35. [35]

    De Gruyter, Berlin/Boston, 2012

    Mark M Meerschaert and Alla Sikorskii.Stochastic Models for Fractional Calculus, volume 43 ofDe Gruyter Studies in Mathematics. De Gruyter, Berlin/Boston, 2012

    work page 2012

  36. [36]

    Statistical models for earthquake occurrences and residual analysis for point processes.Journal of the American Statistical Associa- tion, 83(401):9–27, 1988

    Yosihiko Ogata. Statistical models for earthquake occurrences and residual analysis for point processes.Journal of the American Statistical Associa- tion, 83(401):9–27, 1988

    work page 1988

  37. [37]

    Papaspiliopoulos, M

    O. Papaspiliopoulos, M. Ruggiero, and D. Span` o. Conjugacy properties of time-evolving Dirichlet and gamma random measures.Electronic Journal of Statistics, 10(2):3452–3489, 2016

    work page 2016

  38. [38]

    Optimal filtering and the dual process.Bernoulli, 20(4):1999–2019, 2014

    Omiros Papaspiliopoulos and Matteo Ruggiero. Optimal filtering and the dual process.Bernoulli, 20(4):1999–2019, 2014

    work page 1999

  39. [39]

    Saichev and George Zaslavsky

    A. Saichev and George Zaslavsky. Fractional kinetic equations: Solutions and applications.Chaos (Woodbury, N.Y.), 7:753–764, 01 1998

    work page 1998

  40. [40]

    Springer Series in Operations Research and Financial Engineering

    Gennady Samorodnitsky.Stochastic Processes and Long Range Depen- dence. Springer Series in Operations Research and Financial Engineering. Springer, Cham, Switzerland, 2016

    work page 2016

  41. [41]

    Ewf: simulating exact paths of the wright–fisher diffusion.Bioinformatics, 39(1):btad017, 01 2023

    Jaromir Sant, Paul A Jenkins, Jere Koskela, and Dario Span` o. Ewf: simulating exact paths of the wright–fisher diffusion.Bioinformatics, 39(1):btad017, 01 2023

    work page 2023

  42. [42]

    Sato.L´ evy Processes and Infinitely Divisible Distributions

    K. Sato.L´ evy Processes and Infinitely Divisible Distributions. Cambridge University Press, 1999

    work page 1999

  43. [43]

    Line-of-descent and genealogical processes, and their ap- plications in population genetics models.Theoretical Population Biology, 26(2):119–164, 1984

    Simon Tavar´ e. Line-of-descent and genealogical processes, and their ap- plications in population genetics models.Theoretical Population Biology, 26(2):119–164, 1984

    work page 1984

  44. [44]

    B. Toaldo. Convolution-Type Derivatives, Hitting-Times of Subordinators and Time-ChangedC 0-Semigroups.Potential Analysis, 42:115–140, 2015. 28 Appendix. Subordinated Wright-Fisher priors A-1 Additional plots Figure 1: Laplace transform mapst→E[e −λnC(t)] across the ranget∈[0,2] and for a selection of (n, θ) and subordinators: (a)α-Stable withα= 0.5;(b) P...

    work page 2015

  45. [45]

    We omit the proof and instead refer to the proof of Theorem 4 in [30], of which it follows exactly the same steps, where the weights are constructed from eZ=Z◦Srather thanZ

    It extends the known analogous result on smoothing for WF diffusions, given by Theorem 4 in [30], with the only difference that role of the WF diffusion’s dualZis replaced by the subordinated dual eZ Corollary A-2.The model(41)admits a computable smoothing filter. We omit the proof and instead refer to the proof of Theorem 4 in [30], of which it follows e...

    work page