pith. sign in

arxiv: 2503.03347 · v2 · pith:DMPZVMJJnew · submitted 2025-03-05 · 🧮 math.ST · stat.TH

Drift estimation for rough processes under small noise asymptotic : trajectory fitting method

Arnaud Gloter (LaMME) , Nakahiro Yoshida This is my paper

Pith reviewed 2026-05-23 01:33 UTC · model grok-4.3

classification 🧮 math.ST stat.TH
keywords drift estimationstochastic Volterra equationsmall noise asymptotictrajectory fitting estimatorsingular kernelconsistencyasymptotic normality
0
0 comments X

The pith

A trajectory fitting estimator for the unknown drift in a stochastic Volterra equation with singular kernel is consistent and asymptotically normal as the noise level tends to zero.

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

The paper examines a process X^ε that solves a stochastic Volterra equation whose drift contains an unknown parameter θ* and whose kernel is singular, such as the power-law kernel with exponent α in (0,1/2). The diffusion term is scaled by a small parameter ε that tends to zero. From a single observed path on [0,T], the authors define a trajectory fitting estimator that compares the observed path to the deterministic integral equation driven by candidate drift functions. They prove that this estimator converges to the true parameter and satisfies a central limit theorem, provided identifiability conditions on the drift hold.

Core claim

From an observation of the path (X^ε_s)_{s∈[0,T]}, we construct a Trajectory Fitting Estimator, which is shown to be consistent and asymptotically normal. We also specify identifiability conditions insuring the L^p convergence of the estimator.

What carries the argument

Trajectory Fitting Estimator: the functional that minimizes a distance between the observed path and the solution of the deterministic Volterra equation obtained by replacing the stochastic integral with the candidate drift.

If this is right

  • The estimator converges in probability to the true drift parameter as ε tends to zero.
  • A central limit theorem holds for a suitably normalized version of the estimator.
  • L^p moments of the estimation error converge to zero under the identifiability conditions.
  • The method applies to any singular kernel for which the Volterra equation admits a unique solution.

Where Pith is reading between the lines

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

  • The same fitting construction may extend to other path-dependent or rough-path driven equations whose deterministic skeleton remains well-defined.
  • The asymptotic normality could be used to build confidence intervals for the drift parameter once the asymptotic variance is estimated from the data.

Load-bearing premise

The observed path exactly satisfies the stochastic Volterra equation with the given singular kernel and the diffusion coefficient is exactly proportional to the vanishing parameter ε, while the stated identifiability conditions hold for the true drift.

What would settle it

Numerical paths generated from the Volterra equation with known θ* for successively smaller ε; the estimator should approach θ* at the predicted rate and the properly scaled error should converge in distribution to a normal law.

read the original abstract

We consider a process $X^\ve$ that solves a stochastic Volterra equation with an unknown parameter $\theta^\star$ in the drift function. The Volterra kernel is singular, and includes as an example, $K\_0(u)=c u^{\alpha-1/2} \id{u>0}$ with $\alpha \in (0,1/2)$. It is assumed that the diffusion coefficient is proportional to $\ve \to 0$. From an observation of the path $(X^\ve\_s)\_{s\in[0,T]}$, we construct a Trajectory Fitting Estimator, which is shown to be consistent and asymptotically normal. We also specify identifiability conditions insuring the $L^p$ convergence of the estimator.

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 / 3 minor

Summary. The paper considers a process X^ε solving a stochastic Volterra equation with singular kernel (including the example K_0(u) = c u^{α-1/2} 1_{u>0}, α ∈ (0,1/2)) and drift parameter θ*, where the diffusion coefficient scales with ε → 0. From the continuous observation of the path on [0,T], it defines a trajectory fitting estimator as the minimizer of the integrated squared distance to the deterministic Volterra solution and proves consistency, asymptotic normality, and L^p convergence under explicit identifiability conditions.

Significance. If the results hold, the work extends trajectory-fitting methods to rough Volterra processes with singular kernels under small-noise asymptotics, providing a tractable alternative when likelihood-based methods are difficult. The explicit statement of identifiability conditions and the use of standard M-estimator arguments once uniform convergence of the contrast is shown are strengths that make the claims falsifiable and reproducible in principle.

major comments (2)
  1. [§3] §3, definition of the contrast function (3.2): the integrated squared distance must be shown to be well-defined and finite almost surely for the singular kernel when α < 1/2; the paper needs an explicit integrability check on the difference between the stochastic and deterministic solutions to ensure the minimizer exists.
  2. [§4] §4, Theorem 4.1 (consistency): the proof relies on uniform convergence in probability of the normalized contrast to its deterministic limit; the rate of this convergence should be stated explicitly for the given range of α, as it is load-bearing for the argmin consistency argument.
minor comments (3)
  1. The abstract and §1 should clarify that the L^p convergence holds for p in a specific range rather than all p, to avoid overstatement.
  2. Notation for the deterministic solution x^θ should be introduced earlier and used consistently in the statements of the main theorems.
  3. A short remark comparing the trajectory-fitting estimator to the maximum likelihood estimator (when the latter is tractable) would help situate the contribution.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive evaluation, and constructive suggestions. Both major comments identify points where the manuscript can be strengthened by adding explicit statements; we will incorporate these clarifications in the revision.

read point-by-point responses

  1. Referee: [§3] §3, definition of the contrast function (3.2): the integrated squared distance must be shown to be well-defined and finite almost surely for the singular kernel when α < 1/2; the paper needs an explicit integrability check on the difference between the stochastic and deterministic solutions to ensure the minimizer exists.

    Authors: We agree that an explicit integrability argument is required to rigorously confirm that the contrast (3.2) is well-defined a.s. for α < 1/2. The manuscript implicitly relies on the Hölder regularity of solutions to the Volterra equation, but does not spell out the check. In the revised version we will add a short lemma (placed before Definition 3.2) establishing that, for the given range of α and under the small-noise scaling, X^ε − x^θ belongs to a Hölder space with exponent strictly larger than α almost surely, which guarantees that the integrated squared distance is finite. This will also confirm existence of the minimizer. revision: yes

  2. Referee: [§4] §4, Theorem 4.1 (consistency): the proof relies on uniform convergence in probability of the normalized contrast to its deterministic limit; the rate of this convergence should be stated explicitly for the given range of α, as it is load-bearing for the argmin consistency argument.

    Authors: We acknowledge that stating the explicit rate improves transparency. The proof of Theorem 4.1 invokes uniform convergence in probability of the normalized contrast, which follows from moment bounds on the stochastic convolution term; these bounds yield a rate of order O_p(ε^β) with β = α (for the leading singular term). In the revision we will insert a sentence immediately after the statement of uniform convergence that records this rate and verifies that it is sufficient for the standard argmin consistency argument used in the proof. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper defines a trajectory-fitting estimator as the argmin of an integrated squared contrast between the observed rough path and the deterministic Volterra solution driven by candidate drift parameter θ. Consistency and asymptotic normality as ε→0 are derived from standard M-estimator theory once uniform convergence of the contrast and explicit identifiability conditions are verified. No equation reduces the estimator to a quantity defined in terms of itself, no fitted parameter is relabeled as a prediction, and no load-bearing step rests on a self-citation whose content is itself unverified or circular. The argument is self-contained against external M-estimator benchmarks and the stated identifiability assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Only abstract available; the central claim rests on the existence of a unique solution to the Volterra equation and on the identifiability conditions mentioned but not displayed.

axioms (2)
  • domain assumption The observed process X^ε solves the stochastic Volterra equation with the stated singular kernel and diffusion scaled by ε.
    Stated as the modeling assumption in the abstract.
  • domain assumption Identifiability conditions hold that guarantee L^p convergence of the estimator.
    Invoked in the abstract to obtain the convergence result.

pith-pipeline@v0.9.0 · 5657 in / 1213 out tokens · 22584 ms · 2026-05-23T01:33:27.573419+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Drift estimation for rough processes under small noise asymptotic : QMLE approach

    math.ST 2025-10 unverdicted novelty 6.0

    Constructs QMLE for drift parameter in singular Volterra SDE with small diffusion, proving path reconstruction error O(h^{1/2}) independent of roughness α and yielding efficient estimator as ε→0.

Reference graph

Works this paper leans on

29 extracted references · 29 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    A Perspective on the Numerical Trea tment of Volterra Equations

    Christopher T. H. Baker. “A Perspective on the Numerical Trea tment of Volterra Equations”. In: Journal of Computational and Applied Mathe- matics. Numerical Analysis 2000. Vol. VI: Ordinary Differential Equations and Integral Equations 125.1 (Dec. 15, 2000), 217–249. issn: 0377-0427. doi: 10.1016/S0377-0427(00)00470-2

    work page doi:10.1016/s0377-0427(00)00470-2 2000

  2. [2]

    Modelling Energy Spot Prices by Volatility Modulated L´ evy-driven Volterra Processes

    Ole E. Barndorff-Nielsen, Fred Espen Benth, and Almut E. D. Vera art. “Modelling Energy Spot Prices by Volatility Modulated L´ evy-driven Volterra Processes”. In: Bernoulli 19.3 (Aug. 2013), 803–845. issn: 1350-7265. doi: 10.3150/12-BEJ476

    work page doi:10.3150/12-bej476 2013

  3. [3]

    Volterra Equations with Itˆ o In- tegrals—I

    Marc A. Berger and Victor J. Mizel. “Volterra Equations with Itˆ o In- tegrals—I”. In: Journal of Integral Equations 2.3 (1980), 187–245. issn: 0163-5549. JSTOR: 26164035

    work page 1980

  4. [4]

    Volterra Equations with Itˆ o Inte- grals—II

    Marc A. Berger and Victor J. Mizel. “Volterra Equations with Itˆ o Inte- grals—II”. In: Journal of Integral Equations 2.4 (1980), 319–337. issn: 0163-5549. JSTOR: 26164044

    work page 1980

  5. [5]

    and Hoffmann, Marc and Liu, Yanghui and Rosenbaum, Mathieu and Szymanski, Gr goire , year=

    Carsten H. Chong et al. “Statistical Inference for Rough Volat ility: Central Limit Theorems”. In: The Annals of Applied Probability 34.3 (June 2024), 2600–2649. issn: 1050-5164, 2168-8737. doi: 10.1214/23-AAP2002

    work page doi:10.1214/23-aap2002 2024

  6. [6]

    Statistical Inference for Rough Volat ility: Min- imax Theory

    Carsten H. Chong et al. “Statistical Inference for Rough Volat ility: Min- imax Theory”. In: The Annals of Statistics 52.4 (Aug. 2024), 1277–1306. issn: 0090-5364, 2168-8966. doi: 10.1214/23-AOS2343

    work page doi:10.1214/23-aos2343 2024

  7. [7]

    S tochastic Volterra Equations with Singular Kernels

    W. George Cochran, Jung-Soon Lee, and J¨ urgen Potthoff. “S tochastic Volterra Equations with Singular Kernels”. In: Stochastic Processes and their Applications 56.2 (Apr. 1, 1995), 337–349. issn: 0304-4149. doi: 10.1016/0304-4149(94)00072-2

    work page doi:10.1016/0304-4149(94)00072-2 1995

  8. [8]

    Roughenin g Hes- ton

    Omar El Euch, Jim Gatheral, and Mathieu Rosenbaum. “Roughenin g Hes- ton”. In: SSRN Electronic Journal (2018). issn: 1556-5068. doi: 10.2139/ssrn.3116887

    work page doi:10.2139/ssrn.3116887 2018

  9. [9]

    The characteristic function of rough Heston models

    Omar El Euch and Mathieu Rosenbaum. “The characteristic func tion of rough Heston models”. English. In: Mathematical Finance 29.1 (Sept. 7, 2016), pp. 3–38. issn: 0960-1627. doi: 10.1111/mafi.12173. arXiv: 1609.02108 [q-fin] . Pre-published

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1111/mafi.12173 2016

  10. [10]

    Con- sistent Estimation for Fractional Stochastic Volatility Model under High- Frequency Asymptotics

    Masaaki Fukasawa, Tetsuya Takabatake, and Rebecca West phal. “Con- sistent Estimation for Fractional Stochastic Volatility Model under High- Frequency Asymptotics”. In: Mathematical Finance 32.4 (2022), 1086–1132. issn: 1467-9965. doi: 10.1111/mafi.12354

    work page doi:10.1111/mafi.12354 2022

  11. [11]

    Quantitative Finance , volume =

    Jim Gatheral, Thibault Jaisson, and Mathieu Rosenbaum. “Volatilit y Is Rough”. In: Quantitative Finance 18.6 (June 3, 2018), 933–949. issn: 1469- 7688, 1469-7696. doi: 10.1080/14697688.2017.1393551. 24

    work page doi:10.1080/14697688.2017.1393551 2018

  12. [12]

    The Quadratic Rough Heston Model and the Joint S&P 500/VIX Smile Calibrati on Prob- lem

    Jim Gatheral, Paul Jusselin, and Mathieu Rosenbaum. The Quadratic Rough Heston Model and the Joint S&P 500/VIX Smile Calibrati on Prob- lem. Jan. 6, 2020. arXiv: 2001.01789 [q-fin] . url: http://arxiv.org/abs/2001.01789. Pre-published

    work page arXiv 2020

  13. [13]

    Parametric Inference for Small Variance and Long Time Horizon McKean-Vlasov Diffusion Mod- els

    Valentine Genon-Catalot and Catherine Lar´ edo. “Parametric Inference for Small Variance and Long Time Horizon McKean-Vlasov Diffusion Mod- els”. In: Electronic Journal of Statistics 15.2 (Jan. 2021), 5811–5854. issn: 1935-7524, 1935-7524. doi: 10.1214/21-EJS1922

    work page doi:10.1214/21-ejs1922 2021

  14. [14]

    Parame tric Infer- ence for Discretely Observed Multidimensional Diffusions with Small Dif - fusion Coefficient

    Romain Guy, Catherine Lar´ edo, and Elisabeta Vergu. “Parame tric Infer- ence for Discretely Observed Multidimensional Diffusions with Small Dif - fusion Coefficient”. In: Stochastic Processes and their Applications 124.1 (Jan. 1, 2014), 51–80. issn: 0304-4149. doi: 10.1016/j.spa.2013.07.009

    work page doi:10.1016/j.spa.2013.07.009 2014

  15. [15]

    Parameter Estimation for a Class of Stochastic Differen- tial Equations Driven by Small Stable Noises from Discrete Observat ions

    Long Hongwei. “Parameter Estimation for a Class of Stochastic Differen- tial Equations Driven by Small Stable Noises from Discrete Observat ions”. In: Acta Mathematica Scientia 30.3 (May 2010), 645–663. issn: 02529602. doi: 10.1016/S0252-9602(10)60067-7

    work page doi:10.1016/s0252-9602(10)60067-7 2010

  16. [16]

    Affine V olterra processes

    Eduardo Abi Jaber, Martin Larsson, and Sergio Pulido. “Affine V olterra processes”. English. In: The Annals of Applied Probability 29.5 (Oct. 22, 2019), pp. 3155–3200. issn: 1050-5164. doi: 10.1214/19-AAP1477. arXiv: 1708.08796 [math] . Pre-published

    work page doi:10.1214/19-aap1477 2019

  17. [17]

    Kutoyants

    Yu. Kutoyants. Identification of Dynamical Systems with Small Noise . Dordrecht: Springer Netherlands, 1994. isbn: 978-94-010-4444-8 978-94- 011-1020-4. doi: 10.1007/978-94-011-1020-4

    work page doi:10.1007/978-94-011-1020-4 1994

  18. [18]

    Kutoyants

    Yury A. Kutoyants. Statistical Inference for Ergodic Diffusion Processes . Springer Series in Statistics. London: Springer London, 2004. isbn: 978- 1-84996-906-2 978-1-4471-3866-2. doi: 10.1007/978-1-4471-3866-2

    work page doi:10.1007/978-1-4471-3866-2 2004

  19. [19]

    Gagliardo–Nirenberg Inequ alities and Non-Inequalities: The Full Story

    Petru Mironescu and Ha ¨ ım Brezis. “Gagliardo–Nirenberg Inequ alities and Non-Inequalities: The Full Story”. In: Annales de l’Institut Henri Poincar´ e C 35.5 (Aug. 1, 2018), 1355–1376. issn: 0294-1449. doi: 10.1016/j.anihpc.2017.11.007

    work page doi:10.1016/j.anihpc.2017.11.007 2018

  20. [20]

    Volterra Equations Driven by Semimartingales

    Philip Protter. “Volterra Equations Driven by Semimartingales”. In: The Annals of Probability 13.2 (May 1985), 519–530. issn: 0091-1798, 2168- 894X. doi: 10.1214/aop/1176993006

    work page doi:10.1214/aop/1176993006 1985

  21. [21]

    Pr¨ omel and David Scheffels

    David J. Pr¨ omel and David Scheffels. Stochastic Volterra Equations with H¨ older Diffusion Coefficients. Mar. 28, 2023. doi: 10.48550/arXiv.2204.02648. arXiv: 2204.02648. url: http://arxiv.org/abs/2204.02648. Pre-published

    work page doi:10.48550/arxiv.2204.02648 2023

  22. [22]

    Paracontrolled Distribut ion Ap- proach to Stochastic Volterra Equations

    David J. Pr¨ omel and Mathias Trabs. “Paracontrolled Distribut ion Ap- proach to Stochastic Volterra Equations”. In: Journal of Differential Equa- tions 302 (Nov. 2021), 222–272. issn: 00220396. doi: 10.1016/j.jde.2021.08.031

    work page doi:10.1016/j.jde.2021.08.031 2021

  23. [23]

    Discrete-time sim ulation of stochastic Volterra equations

    Alexandre Richard, Xiaolu Tan, and Fan Yang. “Discrete-time sim ulation of stochastic Volterra equations”. English. In: Stochastic Processes and their Applications 141 (Mar. 7, 2022), pp. 109–138. issn: 0304-4149. doi: 10.1016/j.spa.2021.07.003. arXiv: 2004.00340 [math] . Pre-published. 25

    work page doi:10.1016/j.spa.2021.07.003 2022

  24. [24]

    Small-Diffusion Asympt otics for Discretely Sampled Stochastic Differential Equations

    Michael Sørensen and Masayuki Uchida. “Small-Diffusion Asympt otics for Discretely Sampled Stochastic Differential Equations”. In: Bernoulli 9.6 (Dec. 2003), 1051–1069. issn: 1350-7265. doi: 10.3150/bj/1072215200

    work page doi:10.3150/bj/1072215200 2003

  25. [25]

    Information Criteria f or Small Diffusions via the Theory of Malliavin–Watanabe

    Masayuki Uchida and Nakahiro Yoshida. “Information Criteria f or Small Diffusions via the Theory of Malliavin–Watanabe”. In: Statistical Infer- ence for Stochastic Processes 7.1 (Mar. 1, 2004), 35–67. issn: 1572-9311. doi: 10.1023/B:SISP.0000016462.43348.8f

    work page doi:10.1023/b:sisp.0000016462.43348.8f 2004

  26. [26]

    Existence and Uniqueness of Solutions to Stoch astic Volterra Equations with Singular Kernels and Non-Lipschitz Coefficien ts

    Zhidong Wang. “Existence and Uniqueness of Solutions to Stoch astic Volterra Equations with Singular Kernels and Non-Lipschitz Coefficien ts”. In: Statistics & Probability Letters 78.9 (July 2008), 1062–1071. issn: 01677152. doi: 10.1016/j.spl.2007.10.007

    work page doi:10.1016/j.spl.2007.10.007 2008

  27. [27]

    A Generalized Gro nwall Inequality and Its Application to a Fractional Differential Equation

    Haiping Ye, Jianming Gao, and Yongsheng Ding. “A Generalized Gro nwall Inequality and Its Application to a Fractional Differential Equation” . In: Journal of Mathematical Analysis and Applications 328.2 (Apr. 2007), 1075–1081. issn: 0022247X. doi: 10.1016/j.jmaa.2006.05.061

    work page doi:10.1016/j.jmaa.2006.05.061 2007

  28. [28]

    Stochastic Volterra Equations in Banach Spac es and Stochastic Partial Differential Equation

    Xicheng Zhang. “Stochastic Volterra Equations in Banach Spac es and Stochastic Partial Differential Equation”. In: Journal of Functional Analy- sis 258.4 (Feb. 2010), 1361–1425. issn: 00221236. doi: 10.1016/j.jfa.2009.11.006

    work page doi:10.1016/j.jfa.2009.11.006 2010

  29. [29]

    Volterra Processes and Applications in Financ e

    Elizabeth Zuniga. “Volterra Processes and Applications in Financ e”. PhD thesis. Universit´ e Paris-Saclay, June 17, 2021. 26

    work page 2021