Pith. sign in

REVIEW 4 major objections 5 minor 32 references

Stochastic Compartment Model of Epidemic Spreading in Complex Networks with Mortality and Resetting

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read A stochastic epidemic model with arbitrary sojourn times has a stable endemic state whenever the basic reproduction number exceeds one.

desk verdict A modest simulation-plus-mean-field extension showing resetting raises the effective R0; the analytic core is standard, but the key resetting claim lacks derivation and the validation is partly circular. read the letter →

arxiv 2509.06039 v2 pith:GPZM7NXD submitted 2025-09-07 cond-mat.stat-mech physics.soc-ph

classification cond-mat.stat-mechphysics.soc-ph MSC 92D3005C82
keywords compartmentmodelrandomwalksstochasticresettingendemicequilibriumWatts-StrogatznetworkmortalitybasicreproductionnumberGammasojourntimes
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper builds a stochastic compartment model of epidemic spread in which individuals are random walkers moving on a complex network, passing through susceptible, exposed, infectious, recovered, and dead states with random sojourn times drawn from Gamma distributions. For zero mortality, it proves that when the basic reproduction number R0 equals the infection rate times the mean infectious period and exceeds one, the disease-free state is unstable and a globally stable endemic equilibrium exists, with explicit formulas giving each compartment's fraction as a ratio of mean sojourn times. Random-walk simulations on a Watts-Strogatz network confirm the formulas, and show that stochastic resetting merely rescales the infection rate so that the effective R0 grows with the resetting probability. If true, this means long-range travels can switch a dying outbreak into a persistent endemic disease, and cutting such travels can block spread.

What carries the argument

The central object is the stochastic compartment model written as exact transition equations with delayed, averaged terms, where disease incidence follows mass action A(t) = beta S(t) J(t) and compartment sojourn times are independent Gamma-distributed random variables. Averaging produces convolution-type evolution equations; Laplace transformation and the final-value theorem yield the endemic equilibrium formulas. On the network side, each walker moves by the transition matrix W_{i->j} = q Pi_{i->j} + p R_j, with q = 1 - p, so resetting introduces long-range relocations that in the large-time limit only change the effective beta.

What would settle it

Run random-walk SEIR simulations on a strongly heterogeneous network with Gamma sojourn times and several resetting probabilities, fit beta(p) from the initial growth rate, and compare the long-time endemic fractions to 1/beta(p) and to the mean-sojourn-time ratios; if the susceptible fraction is not close to 1/R0 or the E:I:R ratios deviate from the mean sojourn times, the claim of universal endemic formulas fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that a multiple-random-walker compartment model with independent, arbitrarily distributed sojourn times in the S, E, I, and R compartments retains the classical endemic-equilibrium structure: the endemic susceptible fraction is 1/R0, and the exposed, infectious, and recovered fractions are proportional to their mean sojourn times. The proof passes through exact stochastic transition equations, averages over the random sojourn times, and uses Laplace transforms and the final-value theorem. For zero mortality, the disease-free equilibrium is stable when R0 < 1 and unstable when R0 > 1, with a globally stable endemic state for R0 > 1. Simulations on a Watts-Strogatz

Load-bearing premise

The central claim rests on treating the infection rate as a single constant beta times S(t)J(t), independent of network details except through beta, so that resetting only changes beta; if network structure or resetting alters incidence in a non-multiplicative way, the endemic formulas need not hold.

Editorial extensions

If this is right

  • If R0 is at or below one, the disease dies out; once R0 exceeds one, the disease settles into a globally stable endemic state whose susceptible fraction is exactly 1/R0, independent of initial conditions.
  • The endemic fractions of exposed, infectious, and recovered individuals are proportional to their mean sojourn times, so the ratios of these compartments are set by the disease's biological timing rather than by network details.
  • Stochastic resetting, interpreted as long-range travel, monotonically raises the effective R0, so reducing long-range connections or journeys can push a spreading disease back below threshold.
  • The formulas generalize classical SIR and SEIR endemic results to arbitrary finite-mean sojourn-time distributions, not just memoryless exponential stages.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the mean-field incidence assumption is correct, then any observed mismatch between the endemic ratios and the mean sojourn times would signal non-mass-action effects, such as spatial correlations or network clustering, that the current model does not capture.
  • A testable extension is to fit beta(p) from the early exponential growth in simulations and then predict the full endemic state; if the prediction fails on strongly heterogeneous networks, the infection rate needs a topology-dependent correction beyond a single multiplicative constant.
  • With mortality included, the conservation law S + E + I + R = 1 is broken, so the zero-mortality endemic formulas cannot be directly rescaled by the surviving population; the paper explores mortality numerically but its analytical proof is restricted to zero mortality.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript introduces a stochastic compartment model (SEIRD) with independent Gamma-distributed sojourn times and mortality, coupled to random-walk simulations on a Watts-Strogatz network. For zero mortality, it derives an endemic equilibrium with S_e = 1/R0, E_e, J_e, R_e proportional to the mean sojourn times, where R0 = beta <t_I>, and claims that the disease-free state is unstable and the endemic state is globally stable for R0 > 1. The paper then simulates the random-walk process with stochastic resetting and claims that resetting only renormalizes beta, so that the same endemic formulas hold with R0 monotonically increasing in the resetting probability p. Simulation results are reported for one WS graph with and without mortality.

Significance. If the claims hold, the paper offers a useful bridge between microscopic random-walk dynamics and macroscopic compartment models with non-exponential sojourn times, and it identifies stochastic resetting as a simple mechanism by which long-range travel can raise R0 above unity. The analytical formula (6) is compact and testable, and the E_e:J_e:R_e ratio check is a genuine non-circular corroboration. However, the novel claim about resetting is not currently supported by a derivation or by independent measurement; the stability proof is not displayed; and the main simulation validation is partly circular. The significance is therefore conditional on the missing support being supplied.

major comments (4)
  1. [Section V, Eq. (8)] The assertion that stochastic resetting 'in the large time limit affects only the macroscopic transmission coefficient beta' is load-bearing and unproven. The microscopic transition matrix (8) changes the mixing statistics and pairwise encounter rates; no argument is given that the mass-action incidence (1) with a single constant beta(p) remains valid, even approximately, at large times. The paper's own Conclusion lists 'infection rates beyond the present mass-action law ... including information of the network topology and the random walk' as future work, acknowledging the gap. Without a derivation or an independent measurement of beta(p), the claim that the endemic formulas hold for any resetting rate p is not established.
  2. [Section V, first paragraph, Eq. (6)] The simulation validation is circular for the main quantitative trend. R0 is determined from the measured S_e via the first equation of (6), i.e., S_e = 1/R0. The reported monotone increase of R0 with p is therefore a restatement of the measured S_e(p), not an independent test. The E_e:J_e:R_e ratios do corroborate one consequence of (6), but they do not validate the effective-beta substitution or the absolute endemic values. The text should either measure beta independently (e.g., from the early growth rate or from contact counts) or explicitly limit the claim to a consistency check of the ratios.
  3. [Section III, Eqs. (2)-(6)] The paper states 'we prove the existence of an endemic equilibrium' and that the disease-free state is unstable for R0 > 1, but no proof is displayed. The relevant equations are referenced as '??', and no linearization, Lyapunov function, or Laplace-transform argument is shown. If the proof is a routine generalization of results in refs. [22,23], it should still be sketched here, since the theorem is one of the central claims. As written, the proof is an omitted derivation rather than a presented one.
  4. [Figures 1-2 and Section IV] All simulation conclusions are drawn from single trajectories on one network (1500 nodes, 200 walkers) without multiple-run statistics, error bars, or finite-size analysis. Given that the paper claims 'excellent agreement' and a monotonically increasing R0(p), the absence of uncertainty quantification leaves the quantitative assertions unsupported. Reporting mean and standard deviation over independent realizations, at least for the endemic values, would be needed to assess the agreement.
minor comments (5)
  1. [Throughout] Equation references appear as '??' in many places (e.g., Eqs. (2), (3), (5), (6) and the figure references), making the manuscript not self-contained. Please ensure all cross-references are correct.
  2. [Section III, initial conditions] The notation R(0)=R0 conflicts with the basic reproduction number R0 defined later in the same section. Please rename the initial recovered fraction (e.g., R_initial or R_v) to avoid ambiguity.
  3. [Section II] Typo: 'non-Marlovian' should be 'non-Markovian'.
  4. [Figure captions] Figure 1 states there are 200 walkers, while Figure 2 says 'survived walkers out of 1500'. The population size should be clarified; it appears the number of walkers is 200, not 1500.
  5. [Figure 1 caption] 'large world Watts-Strogatz (WS) network' is likely a typo for 'small-world' network.

Circularity Check

2 steps flagged · score 6.0 of 10

Endemic-state validation under resetting is partly circular: R0 is read off S_e via eq. (6), so the S_e match and the monotone R0(p)/beta(p) trend are fitted rather than predicted; only the E:J:R ratio check gives independent confirmation.

  1. self definitional [Section V (Results and Discussion), paragraph after the first figure]
    "We determined R0 in the simulations from the first equation of (??)."

    The first equation of (6) is S_e = 1/R0. Solving it for R0 from the simulated S_e makes the reported agreement of S_e with 1/R0 tautological. The stated monotone increase of R0 with resetting probability p is therefore just the inverse of the measured decrease of S_e(p), not an independent test of the endemic formula or of the claim that resetting only renormalizes beta.

  2. fitted input called prediction [Section V, paragraph: 'These endemic values are in excellent agreement...']
    "This remains true when the random walks of the individuals are subjected to resetting, which in the large time limit affects only the macroscopic transmission coefficient β."

    β is never measured independently in the simulations; the only inferred quantity is R0 = 1/S_e from eq. (6), equivalently β = R0/<t_I>. The assertion that resetting 'affects only β' is thus a restatement of the fitting step: any p-dependence of the observed S_e is absorbed into β by construction. No microscopic derivation from the resetting transition matrix (8) to a mass-action law with renormalized β is provided, so the transfer of formulas (6) to arbitrary p is not an independently tested prediction.

full rationale

The no-resetting endemic formulas (6) are a legitimate analytic result, although the proof is delegated to the authors' own refs. [22,23]; I do not treat that delegation as circular because those are earlier parameter-free derivations. The circularity enters in the resetting validation. In Section V the authors determine R0 from the simulated endemic susceptible fraction using the first equation of (6), i.e. R0 = 1/S_e. Consequently, the reported agreement of S_e with 1/R0 is tautological, and the asserted monotone increase of R0 with p is the inverse of the measured S_e(p) curve. The only non-tautological content is the E:J:R ratio check and the qualitative time evolution. The further claim that stochastic resetting 'affects only the macroscopic transmission coefficient β' is also unfalsified within the paper: β is not measured independently; it is the fitting parameter that absorbs the p-dependence of S_e via β = R0/<t_I>. The Conclusion acknowledges that infection rates encoding network topology and random-walk structure are future work, confirming that the renormalization step is an assumption rather than a derived microscopic result. Thus the central resetting claim is partially circular, though the underlying no-resetting theory retains independent content.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced; resetting is a known stochastic process and the D compartment is a standard model component. The predictive content rests on input distributions, the mass-action closure, and the assumption that resetting only renormalizes beta.

free parameters (4)
  • beta (infection transmission coefficient) = effective; R0 / <t_I>, not directly measured
    In Eq. (1) A(t) = beta S(t) J(t); beta sets R0 = beta <t_I> and all endemic levels, but its relation to walker contact probability and network structure is not derived.
  • Reset probability p = scanned values including 0, 0.2, 0.6
    Controls long-range jumps in Eq. (8); the paper's main simulation variable; no analytic formula for R0(p) is derived.
  • Compartment sojourn-time distributions (Gamma) = <t_E> : <t_I> : <t_R> = 1 : 4 : 2 in simulations; shape parameters unspecified
    Endemic ratios depend on the means; Gamma shape parameters are inputs, not derived.
  • Mortality survival-time distribution K_M = not specified, simulation labeled 'high mortality'
    Determines deaths through t_M < t_I in Eq. (3); parameters never given.
assumptions (4)
  • domain assumption Mass-action incidence A(t) = beta S(t) J(t) closes the microscopic random-walk contact process
    Invoked in Eq. (1); the paper lists network topology dependent infection rates as future work, so this is load-bearing.
  • domain assumption Compartmental sojourn times t_E, t_I, t_R, t_M are independent with finite means
    Needed for convolution averaging and for the limit-value theorem used to derive the endemic equilibrium in Section III.
  • standard math Laplace final-value theorem and convolution theorem apply to the averaged delayed equations
    Used implicitly to derive the endemic equilibrium from the Laplace-transformed equations in Section III.
  • ad hoc to paper Resetting only renormalizes beta in the large-time limit
    Stated in Section V; this identification is assumed, not derived, and is essential to claim the same endemic formulas hold for all p.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Stochastic Compartment Model of Epidemic Spreading in Complex Networks with Mortality and Resetting." pith.science (2026). https://pith.science/paper/GPZM7NXD

@misc{pith2026250906039,
  author       = {Pith},
  title        = {Pith review of: Stochastic Compartment Model of Epidemic Spreading in Complex Networks with Mortality and Resetting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GPZM7NXD}},
  note         = {Machine review of arXiv:2509.06039}
}
abstract

We propose an epidemic compartment model, which includes mortality caused by the disease, but excludes demographic birth and death processes. Individuals are represented by random walkers, which are in one of the following states (compartments) S (susceptible to infection), E (exposed: infected but not infectious corresponding to the latency period), I (infected and infectious), R (recovered, immune), D (dead). The disease is transmitted with a certain probability at contacts of I to S walkers. The compartmental sojourn times are independent random variables drawn from specific (here Gamma-) distributions. We implement this model into random walk simulations. Each walker performs an independent simple Markovian random walk on a graph, where we consider a Watts-Strogatz (WS) network. Only I walkers may die. For zero mortality, we prove the existence of an endemic equilibrium for basic reproduction number ${\cal R}_0 > 1$ and for which the disease free (globally healthy) state is unstable. We explore the effects of long-range-journeys (stochastic resetting) and mortality. Our model allows for various interpretations, such as certain chemical reactions, the propagation of wildfires, and in population dynamics.

Figures

Figures reproduced from arXiv: 2509.06039 by the authors.

Figure 1
Figure 1. Effect of resetting on the spreading for zero mortality with emergence of endemic states in a large world Watts-Strogatz (WS) network (generated by [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Spreading with high mortality and resetting in the WS graph of Figure [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 32 canonical work pages

  1. [1]

    W. O. Kermack and A. G. McKendrick, A contribution to the mathematical theory of epidemics, Proc. Roy. Soc. A 115, 700–721, 1927

  2. [2]

    H. E. Soper, The interpretation of periodicity in disease prevalence, J. Royal Statistical Society 92, 34-61, 1929

  3. [3]

    W. M. Liu, H. W. Hethcote and S. A. Levin, Dynamical behavior of epidemiological models with non-linear incidence rate. J. Math. Biol. 25, 359–380, 1987

  4. [4]

    M. Y . Li, J. R. Graef, L. Wang and J. Karsai, Global dynamics of a SEIR model with varying total population size. Math. Biosci. 160:191–213, 1999

  5. [5]

    R. M. Anderson and R. M. May, Infectious Diseases in Humans, (Oxford University Press, Oxford), 1992

  6. [6]

    Martcheva, An Introduction to Mathematical Epidemiology, Springer, 2015

    M. Martcheva, An Introduction to Mathematical Epidemiology, Springer, 2015

  7. [7]

    J. E. Harris, Population-Based Model of the Fraction of Incidental COVID-19 Hospitalizations during the Omicron BA.1 Wave in the United States, COVID 3(5), 728-743, 2023. Doi: 10.3390/covid3050054

  8. [8]

    Pastor-Satorras, C

    R. Pastor-Satorras, C. Castellano and P. Van Mieghem, A. Vespignani, Epidemic processes in complex networks, Rev. Mod. Phys. 87, 925-979, 2015

Show all 32 references
  1. [9]

    Pastor-Satorras and A

    R. Pastor-Satorras and A. Vespignani, Epidemic dynamics and endemic states in complex networks, Phys. Rev. E 63, 066117, 2001

  2. [10]

    Okabe Y and A

    Y . Okabe Y and A. Shudo, Microscopic Numerical Simulations of Epidemic Models on Networks. Mathematics 9, 932, 2021

  3. [11]

    Barabási, Network Science

    A.-L. Barabási, Network Science. Cambridge University Press, Cam- bridge, 2016

  4. [12]

    Barabási and R

    A.-L. Barabási and R. Albert, Emergence of Scaling in Random Networks, Science 286, 509, 1999

  5. [13]

    Barrat, M

    A. Barrat, M. Barthélemy and A. Vespignani, Epidemic spreading in population networks, In: Dynamic Processes on Complex Net- works, pp. 180 – 215, Cambridge University Press, 2008, DOI: 10.1017/CBO9780511791383.010

  6. [14]

    Ross, Stochastic Processes (John Wiley & Sons, New York), 1996

    S.M. Ross, Stochastic Processes (John Wiley & Sons, New York), 1996

  7. [17]

    Bestehorn, A

    M. Bestehorn, A. P. Riascos, T. M. Michelitsch and B. A. Collet, A Markovian random walk model of epidemic spreading, Continuum Mech. Thermodyn. 33:1207–1221, 2021. Doi: doi.org/10.1007/s00161- 021-00970-z

  8. [18]

    N. G. van Kampen, Stochastic processes in chemistry and physics (North Holland, Amsterdam), 1981

  9. [19]

    Van Mieghem, Exact Markovian SIR and SIS epidemics on networks and an upper bound for the epidemic threshold, 2014, arXiv:1402.1731

    P. Van Mieghem, Exact Markovian SIR and SIS epidemics on networks and an upper bound for the epidemic threshold, 2014, arXiv:1402.1731

  10. [20]

    Tomovski, T

    L Basnarkov, I. Tomovski, T. Sandev and L. Kocarev, Non-Markovian SIR epidemic spreading model of COVID-19, Chaos, Solitons and Fractals 160, 112286, 2022

  11. [21]

    Bestehorn, T

    M. Bestehorn, T. M. Michelitsch, B. A. Collet, A. P. Riascos, and A. F. Nowakowski, Simple model of epidemic dynamics with memory effects, Phys. Rev. E 105, 024205, 2022

  12. [22]

    Granger, T

    T. Granger, T. M. Michelitsch, M. Bestehorn, A. P. Riascos and B. A. Collet, Four-compartment epidemic model with retarded transition rates, Phys. Rev. E 107 044207, 2023

  13. [23]

    Granger, T

    T. Granger, T. M. Michelitsch, M. Bestehorn, A. P. Riascos and B. A. Collet, Stochastic Compartment Model with Mortality and Its Application to Epidemic Spreading in Complex Networks, Entropy 26(5), 362, 2024

  14. [24]

    Bestehorn and T

    M. Bestehorn and T. M. Michelitsch, Oscillating Behavior of a Compartmental Model with Retarded Noisy Dynamic Infection Rate, Int. J. Bifurcation Chaos 33 2350056, 2023

  15. [25]

    A. P. Riascos and D. P. Sanders, Mean encounter times for multiple random walkers on networks, Phys. Rev. E 103, 042312, 2021

  16. [26]

    Newman, Networks: An Introduction, Oxford University Press, Oxford, 2010

    M.E.J. Newman, Networks: An Introduction, Oxford University Press, Oxford, 2010

  17. [27]

    Noh and H

    J.-D. Noh and H. Rieger, Random walks on complex networks, Phys. Rev. Lett. 92, No. 11, 2004

  18. [28]

    T. M. Michelitsch, A. P. Riascos, B. A. Collet, A. F. Nowakowski and F. C. G. A. Nicolleau, Fractional Dynamics on Networks and Lattices, ISTE/Wiley, London, 2019

  19. [29]

    M. R. Evans and S. N. Majumdar, Diffusion with stochastic resetting, Phys. Rev. Lett. 106, 160601, 2011

  20. [30]

    Michelitsch, G

    T.M. Michelitsch, G. D’Onofrio, F. Polito and A. P. Riascos, Random walks with stochastic resetting in complex networks: A discrete-time approach, Chaos 35, 013119, 2025

  21. [31]

    A. G. Guerrero-Estrada, A.P. Riascos and D. Boyer, Random walks with long-range memory on networks, Chaos 35, 013117, 2025

  22. [32]

    A. Pal, V . Stojkoski and T. Sandev, Random resetting in search problems. In: Target Search Problems, edited by D. Grebekov, R. Metzler, and G. Oshanin ,Springer Nature, 2024, ISBN: 978-3-031- 67801-1 (arXiv:2310.12057)

  23. [33]

    Sandev, A

    T. Sandev, A. Iomin, J. Kurths and L. Kocarev, Shear-driven anomalous diffusion: Memory effects and stochastic resetting, Physics of Fluids 37, 067101, 2025

  24. [34]

    Bestehorn and T.M

    M. Bestehorn and T.M. Michelitsch, Periodic solutions and chaotic attractors of a modified epidemiological SEIS model, Chaos 35, 023104, 2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.