betaFeedback

WorksMathematics

Finite-resolution information about a common clock: alias depth, holding-time loss, and the sampled Ehrenfest chain

Abstract

We study how much Fisher information about a common rate multiplier survives when a reaction system is observed only at discrete times. For a finite known mark law, the first loss at short sampling times is governed by an alias depth hh, the smallest length mismatch among reaction-count fibres: the unrecovered fraction of reaction-count information is 1−Fp(λ)=Cpλh−1+O(λh)1-F_p(\lambda)=C_p\lambda^{h-1}+O(\lambda^h) with an explicit alias coefficient CpC_p. For a fixed irreducible finite generator, holding times add a separate loss: the missing log-clock information is A2t2+O(t3)A_2t^2+O(t^3), or (A3+H)t3+O(t4)(A_3+H)t^3+O(t^4) when A2=0A_2=0. As a worked example we treat the stationary symmetric Ehrenfest chain of known size Ω\Omega observed at its endpoints, with sampling time tt and λ=Ωt\lambda=\Omega t. For every population and every finite positive tt, the exact endpoint information lies between an explicit quadratic lower bound and its matched Gaussian benchmark, Bn≤Rn<g(t)/λB_n\le R_n<g(t)/\lambda, with relative width at most 1/(2λ)1/(2\lambda) measured from BnB_n. The symmetric compound-Poisson benchmark is uniformly relatively accurate on 1≤λ≤LΩ1\le\lambda\le L_\Omega if and only if LΩ=o(Ω)L_\Omega=o(\Omega). A formula-level comparison with the binomial AR(1) literature (Cui–Lund; Weiß–Kim) has not been completed.

Introduction

Sampling a reaction system at discrete times does not reveal the intervening reaction counts or holding times. Both can carry information about a parameter that multiplies every transition rate. We quantify how much of the complete-trajectory Fisher information about such a common clock survives when only the sampled states are observed, and we locate the first loss exactly. The inferred clock is a fixed parameter; no distribution of random external inputs is assumed, and the physical sampling interval Δ\Delta is held fixed under differentiation. With t=θΔt = \theta\Delta the dimensionless sampling time, the information fraction retained is a function of tt, or of λ=Ωt\lambda= \Omega t for a population of size Ω\Omega.

Results. Reaction-count aliases (§ 2). For a finite known mark law with nonzero integer marks, the alias depth hh is the smallest value of max⁡{∣z∣,∣z′∣}\max\{|z|,|z'|\} over pairs of reaction-count vectors z,z′z,z' of different total length with the same increment. Theorem 2.2 gives the loss of the compound-Poisson count experiment as 1−Fp(λ)=Cpλh−1+Op(λh)1 - F_p(\lambda) = C_p\lambda^{h-1} + O_p(\lambda^h) with an explicit positive alias coefficient CpC_p, and Fp≡1F_p \equiv1 when h=∞h = \infty. Corollary 2.3 writes CpC_p in terms of the hh-step mark random walk and evaluates it for nearest-neighbour and simplex mark laws. Two mark laws with identical Gaussian moments can have different exponents (§ 2.1).

Holding-time loss (§ 3). For a fixed finite irreducible generator QQ, Theorem 3.2 classifies the missing log-clock information MQ(t)\mathcal{M}_Q(t) between the complete trajectory and its endpoints: MQ(t)=A2(Q)t2+OQ(t3)\mathcal{M}_Q(t) = A_2(Q)t^2 + O_Q(t^3), and MQ(t)=[A3(Q)+H(Q)]t3+OQ(t4)\mathcal{M}_Q(t) = [A_3(Q) + H(Q)]t^3 + O_Q(t^4) when A2(Q)=0A_2(Q) = 0. Here A2A_2, A3A_3 are explicit path-alias coefficients and H≥0H \ge0 is the holding-time term, positive exactly when the escape rates are not all equal. Every reversible generator has A2>0A_2 > 0 (Remark 3.3), so the t3t^3 regime requires a generator without two-cycles.

The stationary Ehrenfest chain (§ 4–§ 7) is the worked example, observed at its full endpoint pair for a known population size Ω\Omega. Theorem 6.2 gives, for every Ω\Omega and finite t>0t > 0, the finite Gaussian envelope Bn(t)≤Rn(λ)<g(t)/λB_n(t) \le R_n(\lambda) < g(t)/\lambda for the retained information fraction RnR_n, sharp as Ω→∞\Omega\to\infty at fixed tt. Theorem 7.1 shows that the symmetric compound-Poisson fraction F(λ)F(\lambda) approximates RΩ(λ)R_\Omega(\lambda) uniformly in relative error on 1≤λ≤LΩ1 \le\lambda\le L_\Omega if and only if LΩ=o(Ω)L_\Omega= o(\Omega).

Scope. All statements concern Fisher information, not the efficiency of a particular estimator. The alias and holding-time expansions hold for known finite mark laws, finite constant-escape chains and fixed finite irreducible generators; their remainders are not uniform over increasing populations or disappearing edges. For the Ehrenfest chain, the Gaussian envelope bounds the exact endpoint information at every population and positive sampling time, whereas the sharp window concerns the compound-Poisson fraction. Approximation accuracy does not mean that the retained fraction of trajectory information is close to one. The constants in Theorems 5.2 and 7.1 are universal but not optimized, and the Gaussian envelope is proved for the stationary symmetric Ehrenfest experiment only.

Status of the comparison with binomial autoregression. A formula-level comparison of the Ehrenfest results below with Cui and Lund [8] and Weiß and Kim [34, 33] (binomial AR(1)) has not been completed. The paired aggregate counts of the Ehrenfest chain belong to the classical bivariate-binomial and binomial-autoregressive families, so parts of the Ehrenfest results may already follow from moment calculations in that literature; we have not checked which. No priority over those articles is claimed for the Ehrenfest results.

Relation to previous work. Duval and Hoffmann [13, Definition 1 and Theorems 1–3] studied intensity inference for the same limiting compound-Poisson process with symmetric Bernoulli marks in microscopic, intermediate and macroscopic sampling regimes. Their published increment law and score [13, Eqs. (18) and (22)–[23]] give the benchmark FF used here. Their estimator comparison [13 Theorems 4–5] is relative to the sampled-data experiment, not to a continuously observed jump trajectory. Further coarse-scale limitations for compound-Poisson inference appear in [12]; high-frequency nonparametric estimation of jump laws [7] and high-frequency Fisher-information asymptotics for Lévy models [1] address different experiments.

Likelihood computation for discretely observed finite-state chains is established [18, 5]; Kalbfleisch and Lawless obtain the score and expected information of panel-observed chains from derivatives of the transition matrix. For one-parameter birth or death processes, the choice of observation times and the comparison of discrete with continuous sampling have been treated as optimal-design problems [4, 25], and Aït-Sahalia and Mykland [2] compare the losses caused by discrete and by random sampling of diffusions. We have not compared their small-interval behaviour with Theorem 3.2. Our use of the word alias refers to unequal-length reaction-count fibres. It is distinct from the aliasing and identification problem for continuous-time models observed at discrete times [27], and from the embedding problem of whether, and how uniquely, a transition matrix is the exponential of a generator [19, 17]. Here the generator is known up to the common clock, and the question is how much Fisher information about that clock the sampled states retain. The conditional-score identity used below is the missing-information principle [22, 28]. Endpoint-conditioned jump counts, occupation-time moments, and their use in observed information are established for parameterized finite chains [15, 16, 32]. For reaction networks, see pathwise information [26] and experimental design [20, 14]. Transition-score additivity and Gaussian Fisher-information calculations are established ingredients for correlated processes [30]; the standard Gaussian mean-and-covariance formula [30 Section 3.1] supplies our Gaussian benchmark. Diaconis and Griffiths [10 Theorem 1 and Section 6] characterize correlated Bernoulli pair representations and the associated Ehrenfest transition kernels, including their normal and Poisson process limits; the stationary endpoint representation used here is a classical special case. The quadratic lower projection uses classical score-projection principles and fourth moments.

Moor et al. [24] already demonstrate Gaussian reaction-versus-state information discrepancies and their connection to multistate sensing. Kinetic and clock uncertainty relations [9, 29, 23] identify dynamical activity, timing information, and sufficient count statistics. Integer fibres and moves are standard in algebraic statistics [11], and exact cancellation of lattice jumps and its effect on compound-Poisson intensity estimation are explicit in Coca [6 Section 2.3.1, Eq. (2.17), and Lemma 2.2], whose jump law is unknown. Here the finite mark law is known, and we identify the first unequal-length fibre and its parametric Fisher-information loss coefficient.

Reaction-count aliases determine the first loss term

Let ν1,…,νr∈Zd∖{0}\nu_1,\ldots,\nu_r \in\mathbb{Z}^d \setminus\{0\} be a finite mark set, with pi>0p_i > 0, ∑ipi=1\sum_i p_i = 1. Repeated marks are allowed. Write SS for the matrix with these columns and let

Zi∼indPois⁡(λpi),K=∑iZi,D=SZ,Fp(λ)=1−λ−1EVar⁡(K∣D).(1)Z_i \mathrel{\overset{\mathrm{ind}}{\sim}} \operatorname{Pois}(\lambda p_i), \qquad K = \sum_i Z_i, \qquad D = SZ, \qquad F_p(\lambda) = 1-\lambda^{-1}\mathbb{E}\operatorname{Var}(K\mid D). \tag*{(1)}

This is the frozen-intensity experiment; it is not automatically the exact one-bin experiment of a state-dependent network.

Definition 2.1 (Unequal-length alias depth). For d∈Zdd \in\mathbb{Z}^d and k≥0k \ge0 put

bd,k=∑z∈Nr:Sz=d,∣z∣=k∏i=1rpizizi!.(2)b_{d,k}=\sum_{\substack{z\in\mathbb{N}^r:S z=d,\,\lvert z\rvert=k}}\prod_{i=1}^r\frac{p_i^{z_i}}{z_i!}. \tag*{(2)}

If the fibre over dd supports at least two distinct lengths, denote its smallest and next smallest lengths by mdm_d and kdk_d. Let h=min⁡dkdh=\min_d k_d, or h=∞h=\infty if no such fibre exists.

Equivalently, hh is the smallest value of max⁡{∣z∣,∣z′∣}\max\{\lvert z\rvert,\lvert z'\rvert\} for Sz=Sz′Sz=Sz' and ∣z∣≠∣z′∣\lvert z\rvert\ne\lvert z'\rvert. N\mathbb{N} denotes the nonnegative integers. Same-length aliases do not hide the total reaction count and hence do not lose information about its common intensity.

Theorem 2.2 (Alias-depth expansion). Let FpF_p be the retained fraction of the frozen experiment (1), and let hh, bd,kb_{d,k}, mdm_d, kdk_d be as in the preceding definition. If h<∞h<\infty, then h≥2h\ge2 and

1−Fp(λ)=Cpλh−1+Op(λh),Cp=∑d:kd=h(h−md)2bd,h>0.(3)1-F_p(\lambda)=C_p\lambda^{h-1}+O_p(\lambda^h), \qquad C_p=\sum_{d:k_d=h}(h-m_d)^2b_{d,h}>0. \tag*{(3)}

If h=∞h=\infty, then Fp(λ)=1F_p(\lambda)=1 for every λ>0\lambda>0.

The proof of Theorem 2.2 is in Appendix § 10.

The coefficient has a random-walk form. Let ξ1,ξ2,…\xi_1,\xi_2,\ldots be independent with P(ξℓ=ν)=∑i:νi=νpi\mathbb{P}(\xi_\ell=\nu)=\sum_{i:\nu_i=\nu}p_i, put ζk=ξ1+⋯+ξk\zeta_k=\xi_1+\cdots+\xi_k (ζ0=0\zeta_0=0), and let Zk={d:P(ζk=d)>0}Z_k=\{d:\mathbb{P}(\zeta_k=d)>0\}.

Corollary 2.3 (Random-walk form of the alias coefficient). If h<∞h<\infty, then

Cp=1h!∑m=0h−1(h−m)2P(ζh∈Zm).(4)C_p=\frac{1}{h!}\sum_{m=0}^{h-1}(h-m)^2\mathbb{P}(\zeta_h\in Z_m). \tag*{(4)}

For every finite nonzero mark law, 1−Fp(λ)=[2P(ζ2=0)+12P(ζ2∈Z1)]λ+Op(λ2)1-F_p(\lambda)=\left[2\mathbb{P}(\zeta_2=0)+\frac{1}{2}\mathbb{P}(\zeta_2\in Z_1)\right]\lambda+O_p(\lambda^2), and the bracket vanishes exactly when h≥3h\ge3. In particular:

  1. for nearest-neighbour marks ±ej\pm e_j in Zd\mathbb{Z}^d with probabilities pj+,pj−≥0p_j^+,p_j^-\ge0 (marks of probability zero omitted), one has h=2h=2 if some pj+pj−>0p_j^+p_j^->0, and then Cp=4∑jpj+pj−C_p=4\sum_jp_j^+p_j^-; otherwise h=∞h=\infty;

  1. for the simplex marks e1,…,ed,−(e1+⋯+ed)e_1,\ldots,e_d,-(e_1+\cdots+e_d) in Zd\mathbb{Z}^d, d≥1d\ge1, with probabilities p1,…,pd,p0>0p_1,\ldots,p_d,p_0>0, one has h=d+1h=d+1 and Cp=(d+1)2∏i=0dpiC_p=(d+1)^2\prod_{i=0}^d p_i.

References

  1. [1]Y. Aït-Sahalia and J. Jacod. “Fisher’s Information for Discretely Sampled Lévy Processes”. In: Econometrica 76.4 (2008), pp. 727–761. DOI: 10.1111/j.1468-0262.2008.00858.x.
  2. [2]Y. Aït-Sahalia and P. A. Mykland. “The effects of random and discrete sampling when estimating continuous-time diffusions”. In: Econometrica 71.2 (2003), pp. 483–549. DOI: 10.1111/1468-0262.t01-1-00416.
  3. [3]D. F. Anderson, G. Craciun, and T. G. Kurtz. “Product-form stationary distributions for deficiency zero chemical reaction networks”. In: Bulletin of Mathematical Biology 72.8 (2010), pp. 1947–1970. DOI: 10.1007/s11538-010-9517-4. arXiv: 0803.3042.
  4. [4]G. Becker and G. Kersting. “Design problems for the pure birth process”. In: Advances in Applied Probability 15.2 (1983), pp. 255–273. DOI: 10.2307/1426436.
  5. [5]M. Bladt and M. Sørensen. “Statistical inference for discretely observed Markov jump processes”. In: Journal of the Royal Statistical Society. Series B (Statistical Methodology) 67.3 (2005), pp. 395–410. DOI: 10.1111/j.1467-9868.2005.00508.x.DOI
  6. [6]A. J. Coca. “Efficient nonparametric inference for discretely observed compound Poisson processes”. In: Probability Theory and Related Fields 170 (2018), pp. 475–523. DOI: 10.1007/s00440-017-0761-5. arXiv: 1512.08472.DOI
  7. [7]F. Comte and V. Genon-Catalot. “Nonparametric adaptive estimation for pure jump Lévy processes”. In: Annales de l’Institut Henri Poincaré, Probabilités et Statistiques 46.3 (2010), pp. 595–617. DOI: 10.1214/09-AHP323.DOI
  8. [8]Y. Cui and R. Lund. “Inference in binomial AR(1) models”. In: Statistics & Probability Letters 80.23–24 (2010), pp. 1985–1990. DOI: 10.1016/j.spl.2010.09.003.
  9. [9]I. Di Terlizzi and M. Baiesi. “Kinetic uncertainty relation”. In: Journal of Physics A: Mathematical and Theoretical 52, 02LT03 (2019). DOI: 10.1088/1751–8121/aeee34. arXiv: 1809.06410.DOI
  10. [10]P. Diaconis and R. Griffiths. “Exchangeable pairs of Bernoulli random variables, Krawtchouk polynomials, and Ehrenfest urns”. In: Australian & New Zealand Journal of Statistics 54.1 (2012), pp. 81–101. DOI: 10.1111/j.1467–842X.2012.00654.x.DOI
  11. [11]P. Diaconis and B. Sturmfels. “Algebraic algorithms for sampling from conditional distributions”. In: The Annals of Statistics 26.1 (1998), pp. 363–397. DOI: 10.1214/aos/1030563990.DOI
  12. [12]C. Duval. “When is it no longer possible to estimate a compound Poisson process?” In: Electronic Journal of Statistics 8.1 (2014), pp. 274–301. DOI: 10.1214/14-EJS885.DOI
  13. [13]C. Duval and M. Hoffmann. “Statistical inference across time scales”. In: Electronic Journal of Statistics 5 (2011), pp. 2004–2030. DOI: 10.1214/11-EJS660. arXiv: 1106.1031.DOI
  14. [14]Z. R. Fox and B. Munsky. “The finite state projection based Fisher information matrix approach to estimate information and optimize single-cell experiments”. In: PLOS Computational Biology 15.1, e1006365 (2019). DOI: 10.1371/journal.pcbi.1006365.DOI
  15. [15]A. Hobolth and J. L. Jensen. “Statistical inference in evolutionary models of DNA sequences via the EM algorithm”. In: Statistical Applications in Genetics and Molecular Biology 4.1, Article 18 (2005). DOI: 10.2202/1544–6115.1127.
  16. [16]A. Hobolth and J. L. Jensen. “Summary statistics for endpoint-conditioned continuous-time Markov chains”. In: Journal of Applied Probability 48.4 (2011), pp. 911–924. DOI: 10.1239/jap/1324046009.DOI
  17. [17]R. B. Israel, J. S. Rosenthal, and J. Z. Wei. “Finding generators for Markov chains via empirical transition matrices, with applications to credit ratings”. In: Mathematical Finance 11.2 (2001), pp. 245–265. DOI: 10.1111/1467–9965.00114.DOI
  18. [18]J. D. Kalbfleisch and J. F. Lawless. “The analysis of panel data under a Markov assumption”. In: Journal of the American Statistical Association 80.392 (1985), pp. 863–871. DOI: 10.1080/01621459.1985.10478195.DOI
  19. [19]J. F. C. Kingman. “The imbedding problem for finite Markov chains”. In: Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete 1.1 (1962), pp. 14–24. DOI: 10.1007/BF00531768.DOI
  20. [20]M. Komorowski, M. J. Costa, D. A. Rand, and M. P. H. Stumpf. “Sensitivity, robustness, and identifiability in stochastic chemical kinetics models”. In: Proceedings of the National Academy of Sciences 108.21 (2011), pp. 8645–8650. DOI: 10.1073/pnas.1015814108.DOI
  21. [21]T. G. Kurtz. “Solutions of ordinary differential equations as limits of pure jump Markov processes”. In: Journal of Applied Probability 7.1 (1970), pp. 49–58. DOI: 10.2307/3212147.DOI
  22. [22]T. A. Louis. “Finding the observed information matrix when using the EM algorithm”. In: Journal of the Royal Statistical Society: Series B (Methodological) 44.2 (1982), pp. 226–233. DOI: 10.1111/j.2517–6161.1982.tb01203.x.DOI
  23. [23]K. Macieszczak. Ultimate Kinetic Uncertainty Relation and Optimal Performance of Stochastic Clocks. Preprint. 2024. DOI: 10.48550/arXiv.2407.09839. arXiv: 2407.09839.DOI
  24. [24]A.-L. Moor, A. Tjalma, M. Reinhardt, P. R. ten Wolde, and C. Zechner. “State- versus Reaction-Based Information Processing in Biochemical Networks”. In: Physical Review Letters 136, 248401 (2026). DOI: 10.1103/nfk6–8x5s. arXiv: 2505.13373.DOI
  25. [25]D. E. Pagendam and P. K. Pollett. “Locally optimal designs for the simple death process”. In: Journal of Statistical Planning and Inference 140.11 (2010), pp. 3096–3105. DOI: 10.1016/j.jspi.2010.04.017.
  26. [26]Y. Pantazis, M. A. Katsoulakis, and D. G. Vlachos. “Parametric sensitivity analysis for biochemical reaction networks based on pathwise information theory”. In: BMC Bioinformatics 14, 311 (2013). DOI: 10.1186/1471–2105–14–311.arxiv.org/abs/1304.3962
  27. [27]P. C. B. Phillips. “The problem of identification in finite parameter continuous time models”. In: Journal of Econometrics 1.4 (1973), pp. 351–362. DOI: 10.1016/0304–4076(73)90021–3.DOI
  28. [28]D. Pollard. “A note on insufficiency and the preservation of Fisher information”. In: From Probability to Statistics and Back: High-Dimensional Models and Processes—A Festschrift in Honor of Jon A. Wellner. Vol. 9. Institute of Mathematical Statistics Collections. Institute of Mathematical Statistics, 2013, pp. 266–275. DOI: 10.1214/12-IMSCOLL919. arXiv: 1107.3797. URL: https://arxiv.org/abs/1107.3797.
  29. [29]K. Prech, G. T. Landi, F. Meier, N. Nurgalieva, P. P. Potts, R. Silva, and M. T. Mitchison. “Optimal Time Estimation and the Clock Uncertainty Relation for Stochastic Processes”. In: Physical Review X 15, 031068 (2025). DOI: 10.1103/prl5-mp8z. arXiv: 2406.19450.DOI
  30. [30]M. Radaelli, G. T. Landi, K. Modi, and F. C. Binder. “Fisher information of correlated stochastic processes”. In: New Journal of Physics 25, 053037 (2023). DOI: 10.1088/1367-2630/acd321.DOI
  31. [31]K. Takeuchi and A. Takemura. “On sum of 0–1 random variables. I. Univariate case”. In: Annals of the Institute of Statistical Mathematics 39.1 (1987), pp. 85–102. DOI: 10.1007/BF02491451. URL: https://www.ism.ac.jp/editsec/aism/pdf/039_1_0085.pdf.DOI
  32. [32]P. Tataru and A. Hobolth. “Comparison of methods for calculating conditional expectations of sufficient statistics for continuous time Markov chains”. In: BMC Bioinformatics 12, 465 (2011). DOI: 10.1186/1471-2105-12-465.DOI
  33. [33]C. H. Weiß and H.-Y. Kim. “Binomial AR(1) processes: moments, cumulants, and estimation”. In: Statistics 47.3 (2013), pp. 494–510. DOI: 10.1080/02331888.2011.605893.
  34. [34]C. H. Weiß and H.-Y. Kim. “Parameter estimation for binomial AR(1) models with applications in finance and industry”. In: Statistical Papers 54 (2013), pp. 563–590. DOI: 10.1007/s00362-012-0449-y.DOI

Paper details

vlnc:W_bqfkss2c
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

OpenAI ChatGPT GPT-5.6 Sol, OpenAI Codex GPT-5.6 Sol, Anthropic Claude Code Claude Opus 5.5

View statement
Looking for related papers…
Contents