{"id":"82644851-59f6-4dcd-9f9c-73bcbeb1f900","arxiv_id":"2412.00405","paper_version":2,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A stochastic SEIQR epidemic model with a generalized incidence function is analyzed for boundedness, permanence, extinction and geometric ergodicity, but the ergodicity proof is not valid as written.","lead":"This paper adds random noise to a five-compartment epidemic model and claims to prove long-run properties: extinction or persistence and geometric convergence to a stationary distribution. The extinction and permanence arguments are mostly standard, but the geometric ergodicity proof relies on an ellipticity assumption that fails at the boundary, so the headline claim is not established.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6.1's proof is invalid: model (2) is not uniformly elliptic, so minorization (P1) from Lemma 2 is unproven; the ergodicity claim lacks support.","rationale":"The paper's central advertised result is V-geometric ergodicity (Theorem 6.1), and the proof has two connected steps: a Lyapunov inequality (P2), which is derived, and a minorization condition (P1), which is disposed of with the assertion that the model is uniformly elliptic. That assertion is false for the multiplicative diagonal noise used in (2), and the transition-kernel check from E=I=0 shows the claimed Lebesgue-based minorization cannot hold on the stated state space. This is not a disagreement with a prevailing consensus; it is an internal failure of the proof of the strongest claim. The reader's weakest_assumption identified exactly this point, and I agree. Other gaps (Theorem 3.1 stated without proof, the misuse of Lemma 1(b) on bounded integrands, the sigma_1 vs sigma_2 typo in Theorem 5.1's condition, missing parameters for Figure 3) reinforce the rejection, but the uniform-ellipticity/minorization failure is the load-bearing one. I would keep the reader's REJECT verdict: as written, the paper does not establish its headline ergodicity theorem.","tokens_in":13187,"tokens_out":8586,"duration_ms":84607,"concrete_test":"Take X0=(0,0,0,0,0) and T>0. By inspection of (2), if E0=I0=0 then dE(t)=dI(t)=0 identically, so P_T(X0, {E>0 or I>0})=0. Since the proof in Theorem 6.1 takes v to be normalized Lebesgue measure on R5_+, v({E>0 or I>0})>0, contradicting the minorization bound P_T(X0,A) >= alpha v(A). Running this check (or simply locating the uniform-ellipticity hypothesis in Proposition 11.1 of [28] and checking diag(sigma_i^2 x_i^2) against it) settles the concern: the ergodicity proof collapses as written.","verdict_should_be":"REJECT","load_bearing_attack":"The decisive step in the proof of Theorem 6.1 is the sentence 'Since model (2) is uniformly elliptic' followed by Proposition 11.1 of [28], used to obtain positive transition densities and hence the minorization condition (P1). This step is false. The diffusion matrix of (2) is diag(sigma_1^2 S^2, ..., sigma_5^2 R^2), which vanishes on every coordinate hyperplane and therefore admits no uniform lower bound on bounded sets intersecting the boundary of R5_+. Proposition 11.1 cannot be invoked to produce densities bounded below on the compact sets used in the proof. The failure is not merely technical: taking X0 with E0=I0=0, equations (2) give dE(t)=0 and dI(t)=0 for all t (drift and diffusion coefficients vanish identically on that set), so the transition kernel has zero probability of reaching any set with positive measure on {E>0 or I>0}. The proof's candidate measure v is normalized Lebesgue measure, so any such A has v(A)>0 and P_T(X0,A) >= alpha v(A) fails. Thus condition (P1) is not proven and Theorem 6.1, the paper's headline contribution, is unsupported. The Lyapunov condition (P2) is fine; the missing piece is exactly the minorization needed for V-geometric ergodicity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a stochastic SEIQR epidemic model with multiplicative environmental noise and a generalized incidence function. It claims four main results: existence and uniqueness of a global positive solution, stochastic ultimate boundedness and permanence, an explicit stochastic extinction threshold, and V-geometric ergodicity of the solution process. The proofs use Lyapunov functions, Itô calculus, and a cited ergodic theorem. Numerical simulations with the truncated Milstein method are provided to illustrate extinction and persistence.","tokens_in":13434,"tokens_out":7263,"duration_ms":70024,"significance":"If correct, the paper would provide a complete long-run analysis of a fairly general stochastic SEIQR model, including an explicit extinction threshold and geometric ergodicity, which are useful for epidemic forecasting and simulation. The Lyapunov-function arguments for boundedness, permanence, and extinction are standard in style but are worked out for this model. The main claimed novelty, V-geometric ergodicity, is however not established by the proof given, and the theorem as stated is false on the stated state space. The paper also delivers a transparent set of simulations, but the simulations do not compensate for the unsupported central theorem.","major_comments":[{"comment":"The proof of Theorem 6.1 relies crucially on the assertion that model (2) is uniformly elliptic and on Proposition 11.1 of [28] to obtain a minorization condition. This assertion is false: the diffusion matrix is diag(σ1S, σ2E, σ3I, σ4Q, σ5R), whose entries vanish on every coordinate hyperplane of R^5_+, so no uniform ellipticity holds on any compact set that intersects the boundary. Consequently, condition (P1) of Lemma 2 is not proven. Moreover, for the initial condition E(0)=I(0)=0, equations (2) imply E(t)≡I(t)≡0 for all t, so the transition kernel is singular and there cannot be a unique stationary distribution with full support reached from every X0 ∈ R^5_+. The stated theorem is therefore unsupported and, as stated, false.","section":"Section 6, Theorem 6.1"},{"comment":"The hypothesis of Theorem 5.1 is written as σ_1^2 ∧ σ_3^2 > 4(βh'(0)S̃0 − μ), but the proof and the conclusion both involve σ_2^2 ∧ σ_3^2. Since σ2 is not constrained by the stated hypothesis, the condition does not imply the claimed exponential decay. The proof also invokes Lemma 1(b) for the integrands E/(E+I) and I/(E+I), but Lemma 1(b) is stated for the coordinate processes X_k, not for bounded functions of them; the cited lemma does not directly cover the stochastic integrals in inequality (6).","section":"Section 5, Theorem 5.1"},{"comment":"The proof of Theorem 4.1 applies Itô's formula to V(X(t)) and uses inequality (11) without first defining V in this section; V is only introduced later in Section 6 as N + 1/N. More substantively, the theorem claims validity for all initial values X0 ∈ R^5_+, but V is undefined when N=0, and the proof does not address boundary initial data. Since the model can have identically zero components from zero initial conditions, the theorem's domain needs to be stated precisely and handled separately.","section":"Section 4, Theorem 4.1"},{"comment":"Theorem 3.1 is stated for any initial value in R^5_+, but its proof is omitted and the cited standard argument uses the function Ṽ = S+E+I+Q+R − ln S − ln E − ln I − ln Q − ln R − 5, which is not defined when any coordinate is zero. The theorem is therefore not established for boundary initial data, and this ambiguity propagates to the subsequent results that rely on positivity of the solution.","section":"Section 3, Theorem 3.1"}],"minor_comments":[{"comment":"The notation \"σ_2^2 ∧ σ_3^2/2\" is ambiguous; it should be written as (σ_2^2 ∧ σ_3^2)/2.","section":"Section 5, proof of Theorem 5.1"},{"comment":"The sentence \"for all measure sets A\" and the definition v(A) = Leb(A ∪ (Bω(0))/Leb(Bω(0)) contain typographical errors: the intended set is A ∩ Bω(0), and Bω(0) is not defined as a ball in R^5.","section":"Section 6, proof of Theorem 6.1"},{"comment":"References [2] and [6] are the same publication, and several other references are repeated in slightly different forms; the bibliography should be cleaned up.","section":"References"},{"comment":"The text says the initial values are \"proportions\" with S0=E0=I0=Q0=R0=0.25, but these sum to 1.25, which is inconsistent with a normalized population; the terminology should be adjusted.","section":"Section 7.3"},{"comment":"Figure 2 is described as comparing simulation histograms with \"theoretical probability density functions,\" but no stationary density is derived anywhere in the paper; the theoretical curves are not specified.","section":"Section 7.3"}],"recommendation":"reject","confidential_remarks":"The manuscript's central theorem (V-geometric ergodicity) is not proven and is false as stated on the closed positive orthant. The remaining results are more standard and might be salvageable, but the main advertised contribution would require a substantially new proof and a corrected statement. The heavy self-citation also makes the incremental novelty difficult to assess, though my decision is based on the technical invalidity of the ergodicity argument."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Thanks for the report. I agree with the reader's verdict, and the stress-test note pinpoints the real problem. The paper is a routine application of the Lyapunov-threshold program to a new combination: a five-compartment SEIQR SDE with a general incidence h(I). That combination is technically new, and the boundedness, permanence, and extinction sections are mostly standard and probably correct modulo small fixes.\n\nThe trouble is Theorem 6.1. The proof asserts that model (2) is uniformly elliptic and invokes Proposition 11.1 of [28] to get transition densities and a minorization condition. This is false. The diffusion matrix is diag(sigma1 S, sigma2 E, sigma3 I, sigma4 Q, sigma5 R), which vanishes on every coordinate hyperplane. Worse, if E(0)=I(0)=0, then E and I remain zero forever, so the process starting there never reaches the interior. The unique stationary distribution claim cannot hold on the entire R^5_+. The theorem as stated is not just unproven; it is false.\n\nThe other issues are secondary. Theorem 3.1 is stated with the proof explicitly omitted; that is a missing piece. Theorem 5.1's condition mixes sigma1^2 and sigma2^2 (the proof uses sigma2^2 ∧ sigma3^2, which is what the conclusion needs). Lemma 1(b) is applied to bounded integrands like E/(E+I), which the lemma as stated does not cover; this is fixable with a standard martingale strong law but as written is a gap. The numerics look illustrative, but Figure 3's parameters are not stated, which makes verification difficult.\n\nOverall, the paper is not ready for publication. The most interesting result is the ergodicity claim, and that is exactly what fails. The rest is incremental. I would desk reject this version. If the authors can repair the ergodicity statement—perhaps by restricting to the interior and proving irreducibility with a different argument—it could be worth a second look, but as it stands the load-bearing result is wrong. Not worth a referee's time until then.","headline":"The new SEIQR-SDE combination is fine, but the headline ergodicity theorem assumes away the model's boundary behavior and is false as stated.","tokens_in":13997,"tokens_out":5690,"would_cite":false,"duration_ms":53604,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H10","60J25","92D30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that a noisy SEIQR epidemic model either dies out exponentially or converges to a unique stationary distribution at a geometric rate.","keywords":["stochastic SEIQR model","environmental noise","generalized incidence function","Lyapunov method","stochastic permanence","stochastic extinction","V-geometric ergodicity","multiplicative Brownian noise"],"falsifier":"On nested compact sets approaching a coordinate hyperplane in $\\mathbb{R}^5_+$, compute the smallest eigenvalue of the noise matrix $\\mathrm{diag}(\\sigma_1S,\\ldots,\\sigma_5R)$: it tends to zero, which directly contradicts the uniform ellipticity the proof invokes for condition (P1) of Theorem 6.1.","tokens_in":12937,"feed_emoji":"🦠","tokens_out":9946,"duration_ms":83937,"temperature":0.7,"pith_summary":"This paper studies a stochastic version of the SEIQR compartment model—susceptible, exposed, infectious, quarantined, recovered—in which each population is perturbed by its own Brownian noise, and the infection term is a general nonlinear incidence function $h(I)$. The authors aim to show that the noisy system is well behaved in the long run: it has a unique global positive solution, is stochastically ultimately bounded and permanent, and either the infection dies out exponentially or the process converges to a unique stationary distribution at a geometric rate. Their main quantitative result is a noise threshold: if $\\mu > \\frac12 \\max_i \\sigma_i^2$ and $\\sigma_2^2\\wedge\\sigma_3^2 > 4(\\beta h'(0)\\tilde S_0-\\mu)$, then $E(t)+I(t)\\to 0$ almost surely. The paper's strongest claim is geometric ergodicity of the five-dimensional Markov process, which would justify using long-run simulations and estimators as if the model had reached equilibrium.","feed_headline":"Noise threshold decides whether a stochastic epidemic dies out","feed_subtitle":"The paper proves explicit noise conditions for exponential extinction or a unique stationary distribution.","key_machinery":"Two ingredients carry the argument. The first is the Lyapunov function $V(X)=N+1/N$, which satisfies $\\mathcal{L}V\\le \\varpi-\\mu V$ (inequality (11)); this gives boundedness, permanence, and the drift part of the ergodicity criterion. The second is the minorization condition (P1) of Lemma 2, which the paper derives from the claim that system (2) is uniformly elliptic and from Proposition 11.1 of [28], yielding a transition density $g_t(X_0,Y)$ that is strictly positive and bounded below on bounded sets. For extinction, the key object is the Itô differential of $\\ln(E+I)$, together with the bound $\\limsup_{t\\to\\infty}\\langle S(t)\\rangle \\le b/(\\eta+\\mu+d_3)=\\tilde S_0$, which converts noise intensities into a quantitative extinction threshold.","core_discovery":"The central claim, in the authors' terms, is Theorem 6.1: the Markov process $X(t)=(S,E,I,Q,R)$ defined by the stochastic system (2) is $V$-geometrically ergodic. That means there is a unique stationary distribution $\\pi$ and constants $C,\\Lambda>0$ such that $|\\mathbb{E}g(X(t))-\\pi(g)|\\le C V(X_0)e^{-\\Lambda t}$ for every initial $X_0\\in\\mathbb{R}^5_+$ and every measurable $g$ bounded by $V$, with $V(X)=N+1/N$ and $N=S+E+I+Q+R$. For the same model the paper proves almost-sure non-explosion in $\\mathbb{R}^5_+$, stochastic ultimate boundedness, stochastic permanence, and an extinction theorem: under the noise condition above, $\\limsup_{t\\to\\infty} t^{-1}\\ln(E(t)+I(t))<0$, so the infected classes vanish exponentially while the time averages of $S,Q,R$ approach explicit constants. The proof of ergodicity combines the Lyapunov inequality $\\mathcal{L}V \\le \\varpi -\\mu V$ with a minorization condition obtained from the diffusion's uniform ellipticity.","pith_inferences":["Editorial inference: the uniform ellipticity used to verify condition (P1) is not satisfied by the multiplicative-noise diffusion $\\mathrm{diag}(\\sigma_1S,\\ldots,\\sigma_5R)$ on the boundary of $\\mathbb{R}^5_+$, since the noise vanishes on coordinate hyperplanes; the minorization condition therefore needs a separate argument before Theorem 6.1 is fully supported.","Editorial inference: if the ergodicity gap is closed, the result would justify Markov-chain Monte Carlo and particle-filter style estimators for the stationary distribution of this class of epidemic models.","Editorial inference: the same Lyapunov function $N+1/N$ and the $\\ln(E+I)$ comparison could be tested on extended compartment models, for example with waning immunity or multiple pathogen strains, where one would predict analogous noise thresholds for extinction."],"forward_implications":["Long-run simulation of the stochastic SEIQR model is justified: from any positive initial condition, sample paths converge to one stationary distribution, so time averages over a long horizon estimate stationary means.","The extinction threshold gives a practical noise-based control criterion: increasing the perturbation intensities $\\sigma_2$ or $\\sigma_3$ on the exposed and infectious compartments beyond the stated level makes the disease die out exponentially almost surely.","After extinction, the long-run averages of susceptible, quarantined, and recovered populations are explicitly computable as $b/(\\eta+\\mu+d_3)$, $bd_3/((\\mu+\\tau)(\\eta+\\mu+d_3))$, and $b[\\eta(\\mu+\\tau)+\\tau d_3]/(\\mu(\\eta+\\mu+d_3)(\\mu+\\tau))$.","The results hold for a general incidence function $h(I)$ with $h(0)=0$, $h'(0)>0$, and $h(I)/I$ decreasing, so they cover saturating and other nonlinear transmission forms, not only bilinear incidence."],"supporting_citations":[{"why":"Supplies the non-explosion theorem used to prove the global positive solution.","marker":"[13]"},{"why":"Provides the standard Lyapunov argument for existence and uniqueness of global positive solutions of SDEs.","marker":"[14]"},{"why":"Gives the SEIQR system and the proof template for the global positive solution and boundedness estimates.","marker":"[15]"},{"why":"Supplies Lemma 1's strong-law estimates and the extinction-threshold technique for infected classes.","marker":"[24]"},{"why":"Supplies the Lyapunov drift criterion used as condition (P2) in the ergodicity lemma.","marker":"[26]"},{"why":"Supplies the minorization-plus-drift ergodicity criterion for SDEs with locally Lipschitz coefficients.","marker":"[27]"},{"why":"Proposition 11.1 is used to turn uniform ellipticity into a strictly positive transition density and the minorization condition (P1).","marker":"[28]"}],"fun_headline_variants":["Noise sets the fate: extinction or stable endemic equilibrium","Stochastic epidemics: extinction or exponential mixing","Environmental noise decides: die out or settle into stationary state","Proof: noise threshold leads to exponential decay or unique steady law","How much noise tips an epidemic into extinction or persistence"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of Theorem 6.1 leans on the premise that the noise is uniformly elliptic—random perturbations have a strength bounded away from zero everywhere—but in this model the noise vanishes whenever a compartment is empty, so the minorization condition is not actually established.","fun_headline_variants_meta":{"raw":{"variants":["Noise sets the fate: extinction or stable endemic equilibrium","Stochastic epidemics: extinction or exponential mixing","Environmental noise decides: die out or settle into stationary state","Proof: noise threshold leads to exponential decay or unique steady law","How much noise tips an epidemic into extinction or persistence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000478,"raw_usage":{"total_tokens":2365,"prompt_tokens":939,"completion_tokens":1426,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":555,"completion_tokens_details":{"reasoning_tokens":1348}},"tokens_in":555,"tokens_out":1426,"duration_ms":9693,"temperature":1.0,"reasoning_tokens":1348,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:26:06.599434+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On nested compact sets approaching a coordinate hyperplane in $\\mathbb{R}^5_+$, compute the smallest eigenvalue of the noise matrix $\\mathrm{diag}(\\sigma_1S,\\ldots,\\sigma_5R)$: it tends to zero, which directly contradicts the uniform ellipticity the proof invokes for condition (P1) of Theorem 6.1.","supporting_citations":[{"cited_title":"Stochastic stability of differential eq uations","cited_arxiv_id":null,"evidence_quote":"Supplies the non-explosion theorem used to prove the global positive solution."},{"cited_title":"Mao, Stochastic Differential Equations and Applicat ions, Woodhead, 1997","cited_arxiv_id":null,"evidence_quote":"Provides the standard Lyapunov argument for existence and uniqueness of global positive solutions of SDEs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the SEIQR system and the proof template for the global positive solution and boundedness estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies Lemma 1's strong-law estimates and the extinction-threshold technique for infected classes."},{"cited_title":"Meyn, R.L","cited_arxiv_id":null,"evidence_quote":"Supplies the Lyapunov drift criterion used as condition (P2) in the ergodicity lemma."},{"cited_title":"Mattingly, A.M","cited_arxiv_id":null,"evidence_quote":"Supplies the minorization-plus-drift ergodicity criterion for SDEs with locally Lipschitz coefficients."},{"cited_title":"Athreya, T","cited_arxiv_id":null,"evidence_quote":"Proposition 11.1 is used to turn uniform ellipticity into a strictly positive transition density and the minorization condition (P1)."}],"review_version":1}