pith. machine review for the scientific record. sign in

arxiv: 2604.25368 · v1 · submitted 2026-04-28 · 🧮 math.ST · math.PR· stat.TH

Recognition: unknown

Self-organized regime switching in null-recurrent dynamics

Angelika Rohde, Johannes Brutsche, Sebastian Hahn

Pith reviewed 2026-05-07 14:37 UTC · model grok-4.3

classification 🧮 math.ST math.PRstat.TH
keywords null-recurrent diffusionprofile maximum likelihood estimatorregime switchingoscillating Brownian motionlocal timeminimax optimalitystatistical inferencecoupling
0
0 comments X

The pith

The profile MLE for the switching threshold ρ in a null-recurrent diffusion converges at rate n^{-(1+γ)/2} to a limit given by the arg sup of a doubly stochastic process involving local time.

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

This paper derives the asymptotic behavior of the profile maximum likelihood estimator for the unknown threshold at which volatility switches in a null-recurrent stochastic differential equation. Observations are taken at intervals Δ = n^{-γ} for γ between 0 and 1, covering cases from discrete Markov chains to continuous monitoring. The estimator converges at the rate n to the power of minus (1 plus γ) over 2, which is shown to be minimax optimal. Its limiting distribution is the arg sup over a doubly stochastic drifted Poisson process that depends on the local time of an oscillating Brownian motion. The limit is continuous in the parameters, which permits construction of confidence intervals and other inference procedures.

Core claim

For the diffusion dX_t = σ(X_t) dW_t with σ piecewise constant at ρ, the profile MLE based on n observations spaced by n^{-γ} satisfies that n^{(1+γ)/2} (estimator minus ρ) converges in law to the arg sup of a doubly stochastic drifted Poisson process explicitly involving the local time of the oscillating Brownian motion. This non-standard limit is independent of γ but the rate depends on it. The continuity of the limit with respect to ρ, α and β enables statistical inference. The proof relates the null-recurrent chain's long-term behavior to infill asymptotics by coupling after establishing self-similarity when the process starts at the true ρ.

What carries the argument

The arg sup functional over a doubly stochastic drifted Poisson process that incorporates the local time of oscillating Brownian motion, which serves as the limiting object for the scaled profile MLE of the threshold ρ.

If this is right

  • The rate n^{-(1+γ)/2} is attained uniformly for all γ in [0,1) and is minimax optimal.
  • The limiting distribution depends continuously on ρ, α, β, allowing for consistent estimation of the limit and thus inference.
  • The same limiting process arises in both low-frequency null-recurrent sampling and high-frequency infill sampling.
  • The self-similarity property of the centered process when started at ρ enables the coupling argument to extend the result beyond the artificial initial condition.

Where Pith is reading between the lines

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

  • This technique of coupling self-similar processes could be applied to parameter estimation in other null-recurrent or recurrent Markov models.
  • Potential connections exist to change-point detection in time series with long memory or null-recurrence.
  • The explicit form of the limit might allow simulation-based inference without needing closed-form distributions.

Load-bearing premise

The coupling argument successfully overcomes the artificial assumption that the process starts exactly at the true threshold ρ by relating the long-term null-recurrent dynamics to infill behavior using the strong Markov property.

What would settle it

A simulation study for large n with γ=0 showing that the empirical distribution of the scaled MLE deviates from the distribution of the arg sup process, or an analytical counterexample where the self-similarity fails to hold under the model assumptions.

Figures

Figures reproduced from arXiv: 2604.25368 by Angelika Rohde, Johannes Brutsche, Sebastian Hahn.

Figure 1
Figure 1. Figure 1: Sotheby’s (BID) stock prices, 1988-2012, from Yahoo!Finance. The process switches its volatility regime at a threshold ρ of approximately 22. MSC2020 subject classifications: Primary 62M05, 62E20; secondary 60F05, 60J10. Keywords and phrases: Profile MLE, null-recurrent Markov chain, self-similarity, coupling, diffusion local time. 1 arXiv:2604.25368v1 [math.ST] 28 Apr 2026 view at source ↗
Figure 2
Figure 2. Figure 2: An illustration of different observation schemes. The red lines correspond to γ = 1, the infill asympotics, where no point after T = 1 is observed. The blue dashes correspond to the low frequency observation scheme γ = 0. The green ones correspond to γ = 1/2, where the observations are more dense with growing n, but also the observation window increases. Likewise, the only observation scheme for which the … view at source ↗
Figure 3
Figure 3. Figure 3: Simulation of a path Xρ0 in black for ρ0 = 0 and Xx0 in red with x0 = 1. The coupled process X c is then given as the blue one. The ticks on the x-axis illustrate the sampling schemes for γ = 0 (blue), γ = 1/2 (green) and γ = 1 (red), compare view at source ↗
Figure 4
Figure 4. Figure 4: The function Fnj ,C for C = 7 and n1 = 100 (blue), n2 = 250 (red) and n3 = 500 (orange). The locatio of the corresponding local maxima are depicted on the x-acis in the respective color. (c) For any 0 < δ < 1 and Cn = o( √ n) there exists a constant c = c(δ, C) > 0 and n0 ∈ N such that for every n ≥ n0, Fn,Cn (x) ≤ −c √ n for all x /∈ [1 − δ, 1 + δ]. PROOF. We start with part (a) and first note that Fn,Cn … view at source ↗
Figure 5
Figure 5. Figure 5: An illustration of discrete realization, linearly interpolated. To construct the injection (F.26), each upcrossing (red) is mapped to its successor, except of the last segment. Here, we chose one of the earlier segments above the threshold that has no preceeding upcrossing. Such a segment always exists if there are three consecutive points above the threshold. Using the above construction of an injective m… view at source ↗
read the original abstract

Based on discrete observations $X_0,X_{\Delta},\dots, X_{n\Delta}$ for $\Delta=n^{-\gamma}$ with $\gamma\in [0,1)$ of the null-recurrent dynamic $dX_t = \sigma(X_t)dW_t$ with a Brownian motion $W$ and $\sigma(x)=\alpha\mathbb{1}\{x<\rho\} + \beta\mathbb{1}\{x\geq \rho\}$, we derive rate of convergence and limiting distribution of the profile MLE for $\rho$. This includes low-frequency asymptotics ($\gamma=0$) for which the observations form a null-recurrent Markov chain. The derived non-standard limit is the argsup over a doubly stochastic drifted Poisson process explicitly involving the local time of oscillating Brownian motion. Its dependence on $\rho$ as well as the unknown volatility levels $\alpha$ and $\beta$ is shown to be continuous w.r.t. the topology of weak convergence, enabling statistical inference. Whereas this limit is independent of the sampling frequency, the profile MLE's rate of convergence equals $n^{-(1+\gamma)/2}$ and is proven to be minimax optimal. The surprising idea of the proof of the limit theorem is to relate the long-term behavior of the null-recurrent Markov chain to the infill asymptotics on a fixed time interval. Indeed, in the very special case that $(X_t)_{t\geq 0}$ is started in the true parameter $X_0=\rho_0$, the process $(X_t-\rho_0)_{t\geq 0}$ is shown to possess a desirable distributional self-similarity. On basis of the strong Markov property, the artificial constallation of starting in $\rho_0$ is finally overcome by a coupling argument.

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

1 major / 2 minor

Summary. The paper derives the rate of convergence and limiting distribution of the profile MLE for the threshold ρ in the null-recurrent diffusion dX_t = σ(X_t) dW_t with σ switching between α below ρ and β above ρ. From discrete observations with mesh Δ = n^{-γ} (γ ∈ [0,1)), the estimator converges at rate n^{-(1+γ)/2} to the arg sup of a doubly stochastic drifted Poisson process driven by the local time of an oscillating Brownian motion. The limit is continuous in ρ, α, β (enabling inference), independent of γ, and the rate is minimax optimal. The proof relates long-term null-recurrent behavior to infill asymptotics via distributional self-similarity when X_0 = ρ_0, then uses the strong Markov property and coupling to remove the artificial initial condition.

Significance. If the results hold, this supplies a novel explicit non-standard limit theory for MLE in null-recurrent regime-switching diffusions, a setting relevant to applications with persistent but non-stationary dynamics. The continuity of the limiting object in the parameters and the minimax optimality are concrete strengths that support practical inference procedures. The self-similarity device for mapping null-recurrent chains to infill asymptotics is an inventive technical contribution that strengthens the paper's methodological value.

major comments (1)
  1. [Proof of the limiting distribution (coupling step after self-similarity)] The coupling argument used to extend the limit from the special case X_0 = ρ_0 (via strong Markov property at the hitting time τ_ρ0) is load-bearing for the central claim. Under null-recurrence, τ_ρ0 has infinite mean and heavy tails, so the pre-hitting occupation measure and local-time accumulation are not obviously controlled uniformly in the rescaled processes. This risks an extra random shift or intensity modulation in the doubly stochastic Poisson process, which would alter the location of the arg sup. Please supply the specific tightness or uniform integrability arguments that ensure the limiting distribution is unaffected (see the proof outline following the self-similarity result).
minor comments (2)
  1. [Abstract] The abstract introduces 'oscillating Brownian motion' without a brief definition or reference; adding one would improve accessibility for readers outside the immediate subfield.
  2. [Main results] Notation for the rescaled processes and the doubly stochastic intensity could be made more explicit with a dedicated display equation early in the main results section.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We are grateful to the referee for the careful reading, positive evaluation of the paper's contributions, and the constructive suggestion regarding the coupling argument. We address the major comment below and will revise the manuscript to incorporate additional details.

read point-by-point responses

  1. Referee: [Proof of the limiting distribution (coupling step after self-similarity)] The coupling argument used to extend the limit from the special case X_0 = ρ_0 (via strong Markov property at the hitting time τ_ρ0) is load-bearing for the central claim. Under null-recurrence, τ_ρ0 has infinite mean and heavy tails, so the pre-hitting occupation measure and local-time accumulation are not obviously controlled uniformly in the rescaled processes. This risks an extra random shift or intensity modulation in the doubly stochastic Poisson process, which would alter the location of the arg sup. Please supply the specific tightness or uniform integrability arguments that ensure the limiting distribution is unaffected (see the proof outline following the self-similarity result).

    Authors: We agree that the coupling step is central and requires explicit control of the pre-hitting contribution under null-recurrence. The proof first derives the limiting distribution under the auxiliary initial condition X_0 = ρ_0 by exploiting the distributional self-similarity of the rescaled oscillating Brownian motion. The extension to arbitrary initial conditions proceeds via the strong Markov property at the hitting time τ_ρ0. To ensure the pre-hitting occupation measure and local-time accumulation do not introduce an extra random shift or intensity modulation in the limit, we construct a coupling between the original process after τ_ρ0 and an independent copy started at ρ_0. The difference between the two rescaled local-time processes is shown to converge to zero in probability under the n^{-(1+γ)/2} scaling. This follows from tightness of the family of rescaled pre-hitting local-time measures, obtained via moment bounds that exploit the self-similarity and the fact that the heavy tails of τ_ρ0 are offset by the local-time scaling of the oscillating Brownian motion. Uniform integrability of the relevant functionals is then established using the continuity of the limiting doubly stochastic drifted Poisson process with respect to weak convergence in the parameters ρ, α, β. We will expand the existing proof outline into a dedicated subsection containing these tightness and uniform-integrability arguments in the revised version. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation relies on external stochastic process theory

full rationale

The paper establishes the limiting distribution of the profile MLE first under the special initial condition X_0 = ρ_0 by deriving distributional self-similarity of (X_t - ρ_0) from the SDE, then extends to general starts via the strong Markov property and a coupling argument at the hitting time τ_ρ0. This chain uses standard results on null-recurrent Markov chains, local times of oscillating Brownian motion, and doubly stochastic Poisson processes without reducing the arg sup limit or the n^{-(1+γ)/2} rate to a fitted parameter or self-citation by construction. The continuity of the limit in (ρ, α, β) is shown directly, and minimax optimality is established separately. No self-definitional step, fitted-input prediction, or load-bearing self-citation appears in the outlined proof strategy.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper relies on standard stochastic calculus axioms and introduces no new free parameters or entities; the limit is derived from existing theory.

axioms (2)
  • standard math Properties of Brownian motion and local time
    Used in defining the limiting process.
  • standard math Strong Markov property
    Invoked for the coupling argument.

pith-pipeline@v0.9.0 · 5626 in / 1090 out tokens · 108068 ms · 2026-05-07T14:37:08.965703+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

27 extracted references · 19 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" url if FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if ne...

  2. [2]

    bbook [author] Aliprantis , Charalambos D. C. D. Border , Kim C. K. C. ( 1994 ). Infinite-dimensional analysis . Studies in Economic Theory 4 . Springer-Verlag, Berlin A hitchhiker's guide . 10.1007/978-3-662-03004-2 1321140 bbook

    work page doi:10.1007/978-3-662-03004-2 1994

  3. [3]

    1999 , PAGES =

    bbook [author] Billingsley , Patrick P. ( 1999 ). Convergence of probability measures , second ed. Wiley Series in Probability and Statistics: Probability and Statistics . John Wiley & Sons, Inc., New York A Wiley-Interscience Publication . 10.1002/9780470316962 1700749 bbook

    work page doi:10.1002/9780470316962 1999

  4. [4]

    , R\"ockner , Michael M

    barticle [author] Blanchard , Philippe P. , R\"ockner , Michael M. Russo , Francesco F. ( 2010 ). Probabilistic representation for solutions of an irregular porous media type equation . Ann. Probab. 38 1870--1900 . 10.1214/10-AOP526 2722788 barticle

    work page doi:10.1214/10-aop526 2010

  5. [5]

    Rohde , Angelika A

    barticle [author] Brutsche , Johannes J. Rohde , Angelika A. ( 2026 ). The level of self-organized criticality in oscillating Brownian motion: n -consistency and stable Poisson-type convergence of the MLE . Ann. Appl. Probab. to appear . barticle

    2026

  6. [6]

    barticle [author] Chan , K. S. K. S. ( 1993 ). Consistency and limiting distribution of the least squares estimator of a threshold autoregressive model . Ann. Statist. 21 520--533 . 10.1214/aos/1176349040 1212191 barticle

    work page doi:10.1214/aos/1176349040 1993

  7. [7]

    Hoffmann , Marc M

    barticle [author] Delattre , Sylvain S. Hoffmann , Marc M. ( 2002 ). Asymptotic equivalence for a null recurrent diffusion . Bernoulli 8 139--174 . 1895888 barticle

    2002

  8. [8]

    barticle [author] Ferger , D. D. ( 2015 ). Arginf-sets of multivariate cadlag processes and their convergence in hyperspace topologies . Theory Stoch. Process. 20 13--41 . 3510226 barticle

    2015

  9. [9]

    Nappo , Giovanna G

    barticle [author] Fischer , Markus M. Nappo , Giovanna G. ( 2010 ). On the moments of the modulus of continuity of I t\^o processes . Stoch. Anal. Appl. 28 103--122 . 10.1080/07362990903415825 2597982 barticle

    work page doi:10.1080/07362990903415825 2010

  10. [10]

    McKean , Henry P

    bbook [author] It\^o , Kiyoshi K. McKean , Henry P. H. P. Jr. ( 1965 ). Diffusion processes and their sample paths . Die Grundlehren der mathematischen Wissenschaften Band 125 . Springer-Verlag, Berlin-New York; Academic Press, Inc., Publishers, New York . 199891 bbook

    1965

  11. [11]

    URLhttps://link.springer.com/10.1007/978-3-030-61871-1

    bbook [author] Kallenberg , Olav O. ( 2021 ). Foundations of modern probability , third ed. Probability Theory and Stochastic Modelling 99 . Springer, Cham . 10.1007/978-3-030-61871-1 4226142 bbook

    work page doi:10.1007/978-3-030-61871-1 2021

  12. [12]

    Wellner , Jon A

    barticle [author] Keilson , Julian J. Wellner , Jon A. J. A. ( 1978 ). Oscillating B rownian motion . J. Appl. Probability 15 300--310 . 10.2307/3213403 474526 barticle

    work page doi:10.2307/3213403 1978

  13. [13]

    barticle [author] Kutoyants , Yury A. Y. A. ( 2012 ). On identification of the threshold diffusion processes . Ann. Inst. Statist. Math. 64 383--413 . 10.1007/s10463-010-0318-1 2878912 barticle

    work page doi:10.1007/s10463-010-0318-1 2012

  14. [14]

    bbook [author] Le Gall , Jean-Fran cois J.-F. c. ( 1984 ). Stochastic analysis and applications . Lecture Notes in Mathematics 1095 . Springer-Verlag, Berlin . 10.1007/BFb0099117 777509 bbook

    work page doi:10.1007/bfb0099117 1984

  15. [15]

    Pigato , Paolo P

    barticle [author] Lejay , Antoine A. Pigato , Paolo P. ( 2018 ). Statistical estimation of the oscillating B rownian motion . Bernoulli 24 3568--3602 . 10.3150/17-BEJ969 3788182 barticle

    work page doi:10.3150/17-bej969 2018

  16. [16]

    ( 1983 )

    barticle [author] Lindvall , Torgny T. ( 1983 ). On coupling of diffusion processes . J. Appl. Probab. 20 82--93 . 10.1017/s0021900200096947 688082 barticle

    work page doi:10.1017/s0021900200096947 1983

  17. [17]

    ( 2026 )

    barticle [author] Mazzonetto , Sara S. ( 2026 ). Rates of convergence to the local time of oscillating and skew B rownian motion . Bernoulli 32 567--589 . 10.3150/25-bej1870 5000311 barticle

    work page doi:10.3150/25-bej1870 2026

  18. [18]

    Pigato , Paolo P

    barticle [author] Mazzonetto , Sara S. Pigato , Paolo P. ( 2024 ). Drift estimation of the threshold O rnstein- U hlenbeck process from continuous and discrete observations . Statist. Sinica 34 313--336 . 4683574 barticle

    2024

  19. [19]

    , Strauch , Claudia C

    barticle [author] Rei , Markus M. , Strauch , Claudia C. Trottner , Lukas L. ( 2026 ). Change point estimation for a stochastic heat equation . Ann. Statist. 54 277--299 . 10.1214/25-AOS2567 5039619 barticle

    work page doi:10.1214/25-aos2567 2026

  20. [20]

    and YOR, M

    bbook [author] Revuz , Daniel D. Yor , Marc M. ( 1999 ). Continuous martingales and B rownian motion , third ed. Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] 293 . Springer-Verlag, Berlin . 10.1007/978-3-662-06400-9 1725357 bbook

    work page doi:10.1007/978-3-662-06400-9 1999

  21. [21]

    Chan , Kung-Sik K.-S

    barticle [author] Su , Fei F. Chan , Kung-Sik K.-S. ( 2015 ). Quasi-likelihood estimation of a threshold diffusion process . J. Econometrics 189 473--484 . 10.1016/j.jeconom.2015.03.038 3414915 barticle

    work page doi:10.1016/j.jeconom.2015.03.038 2015

  22. [22]

    Chan , Kung-Sik K.-S

    barticle [author] Su , Fei F. Chan , Kung-Sik K.-S. ( 2017 ). Testing for threshold diffusion . J. Bus. Econom. Statist. 35 218--227 . 10.1080/07350015.2015.1073594 3622833 barticle

    work page doi:10.1080/07350015.2015.1073594 2017

  23. [23]

    bbook [author] van der Vaart , A. W. A. W. ( 1998 ). Asymptotic statistics . Cambridge Series in Statistical and Probabilistic Mathematics 3 . Cambridge University Press, Cambridge . 10.1017/CBO9780511802256 1652247 bbook

    work page doi:10.1017/cbo9780511802256 1998

  24. [24]

    bbook [author] van der Vaart , A. W. A. W. Wellner , Jon A. J. A. ( 2023 ). Weak convergence and empirical processes---with applications to statistics , second ed. Springer Series in Statistics . Springer, Cham . 10.1007/978-3-031-29040-4 4628026 bbook

    work page doi:10.1007/978-3-031-29040-4 2023

  25. [25]

    Yau , Chun Yip C

    barticle [author] Yuan , Gan G. Yau , Chun Yip C. Y. ( 2026 ). Generalized multivariate threshold autoregressive models with linearly partitioned threshold space . Ann. Statist. to appear . barticle

    2026

  26. [26]

    Rohde , Angelika A

    barticle [author] Brutsche , Johannes J. Rohde , Angelika A. ( 2026 ). The level of self-organized criticality in oscillating Brownian motion: n -consistency and stable Poisson-type convergence of the MLE . Ann. Appl. Probab., to appear . barticle

    2026

  27. [27]

    bbook [author] Tsybakov , Alexandre B. A. B. ( 2009 ). Introduction to nonparametric estimation . Springer Series in Statistics . Springer, New York Revised and extended from the 2004 French original, Translated by Vladimir Zaiats . 10.1007/b13794 2724359 bbook

    work page doi:10.1007/b13794 2009