pith. sign in

arxiv: 2605.29907 · v1 · pith:2MNJ5RLYnew · submitted 2026-05-28 · 🧬 q-bio.QM

Stochastic network epidemic model and particle filter: General framework and application to influenza in Japan

Ihtisham Ul Haq , Serge Richard This is my paper

Pith reviewed 2026-06-28 23:53 UTC · model grok-4.3

classification 🧬 q-bio.QM
keywords stochastic epidemic modelparticle filterlattice graphinfluenza forecastingstate estimationparameter inferencedata assimilationJapan influenza data
0
0 comments X

The pith

A particle filter framework applied to a stochastic 2D lattice epidemic model estimates hidden states and parameters from incomplete observations and generates one-week-ahead influenza forecasts for Japan.

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

The paper presents a stochastic model of disease spread on a two-dimensional lattice graph. It develops particle filter techniques to estimate the hidden number of infections and model parameters from either total counts or partial location data. These methods are validated on synthetic data before being applied to actual influenza reports from Japanese prefectures, including one-week-ahead predictions. A sympathetic reader would care because this offers a way to monitor and predict epidemics using the incomplete data typically available to health authorities.

Core claim

The authors introduce a two-dimensional lattice graph model for infectious disease spread and propose a particle filter based data assimilation framework for the sequential estimation of both model states and unknown parameters. Two methodologies are developed based on the number of infected agents and partial spatial information. The first method is applied to influenza data from Japan, demonstrating effectiveness for monitoring and forecasting.

What carries the argument

The particle filter data assimilation framework for estimating states and parameters in the stochastic lattice epidemic model.

If this is right

  • Enables real-time epidemic monitoring using partial observations.
  • Supports one-week-ahead forecasting simulations from current weekly data.
  • Facilitates adaptive public health decision-making based on estimated states.
  • Handles both aggregate infected counts and partial spatial location information.

Where Pith is reading between the lines

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

  • The approach could be tested on surveillance data from other countries or diseases with similar reporting structures.
  • Spatial estimates from the partial-information method might guide localized interventions if validated further.
  • Longer forecast horizons or integration with mobility data could be explored as extensions of the current one-week setup.

Load-bearing premise

The particle filter methods can reliably estimate both states and parameters in the stochastic lattice model when applied to real-world incomplete data.

What would settle it

A systematic mismatch between the one-week-ahead forecasts generated by the model and the actual subsequent weekly influenza cases in the Japanese prefectures would indicate the framework does not perform as claimed.

Figures

Figures reproduced from arXiv: 2605.29907 by Ihtisham Ul Haq, Serge Richard.

Figure 1
Figure 1. Figure 1: Spread of infection on a 2D lattice graph-based model with transmission rate [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Workflow of the PF methods. Both methods begin with same prior sampling and the key difference is [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Behavior of the model parameter estimated using PF methods. The black curve represents the true [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Posterior distributions of the transmission parameter [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: PF method 1 estimation compared with the observed data. The algorithm is applied using window sizes [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: PF method 2 estimation compared with the observed data. The algorithm is applied using window sizes [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Time snapshots of the distribution of parameters of the selected particles for method 1 for a window size [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Time snapshots of the distribution of parameters of the selected particles for method 2 for a window size [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Box plot for the relative absolute error between the estimated forecast and the observed infected agents [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Dynamics of influenza infections in Okinawa estimated using the PF method 1. The estimated red curve [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Dynamics of influenza infections in Niigata estimated using the PF method 1. The estimated red curve [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Dynamics of influenza infections in kyoto estimated using the PF method 1. The estimated red curve [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Dynamics of influenza infections in Fukui estimated using the PF method 1. The estimated red curve [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: One week ahead model forecast of weekly influenza cases in Okinawa. [PITH_FULL_IMAGE:figures/full_fig_p024_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: One week ahead model forecast of weekly influenza cases in Niigata. [PITH_FULL_IMAGE:figures/full_fig_p024_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: One week ahead model forecast of weekly influenza cases in Kyoto. [PITH_FULL_IMAGE:figures/full_fig_p025_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: One week ahead model forecast of weekly influenza cases in Fukui. [PITH_FULL_IMAGE:figures/full_fig_p026_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: One week ahead model forecast distribution and posterior distribution of particle filter of weekly influenza [PITH_FULL_IMAGE:figures/full_fig_p026_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Estimated transmission rate βˆ t for the graph-based stochastic model in four prefectures of Japan from July 2024 to December 2025: Okinawa, Niigata, Kyoto, and Fukui. The solid curve show the estimated transmission rate values. The shaded regions show the confidence intervals, with blue indicating the 95% confidence interval and sky blue indicating the 65% confidence interval. In order to illustrate the … view at source ↗
Figure 20
Figure 20. Figure 20: Selected and unselected particles at different weeks using the absolute distance criterion. Green represents [PITH_FULL_IMAGE:figures/full_fig_p028_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Estimated transmission rate βˆ t for the deterministic model for four prefectures in Japan from July 2024 to December 2025. The blue curves show the estimated transmission rate values. The shaded regions illustrate the confidence intervals, with blue indicating the 95% interval confidence interval. For comparison, the observed infected cases are also represented in these figures. 30 [PITH_FULL_IMAGE:figu… view at source ↗
read the original abstract

Parameter inference and state estimation in stochastic and partially observed biological systems remain major problems in mathematical biology. In this work, we introduce a two-dimensional lattice graph model for the spread of infectious diseases. Estimating states and parameters in graph-based stochastic epidemic systems is particularly challenging because of randomness and incomplete observations. To address these issues, we propose a particle filter based data assimilation framework for the sequential estimation of both model states and unknown parameters. Two methodologies are developed: one based on the number of infected agents and another based on partial spatial location's information of infected agents on a two-dimensional lattice. The performance of the two methods are firstly analyzed and validated using synthetic data, and the first method is then applied to influenza data collected from different prefectures in Japan between July 2024 and December 2025. One-week-ahead forecasting simulations are also performed using current weekly data. The findings highlight the effectiveness of the proposed PF framework for real-time epidemic monitoring, forecasting, and adaptive public health decision-making.

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 / 1 minor

Summary. The paper introduces a two-dimensional lattice graph stochastic epidemic model and develops a particle filter data assimilation framework for sequential state and parameter estimation. Two variants are presented (one using infected counts, one using partial spatial information on the lattice). The methods are validated on synthetic data and the first variant is applied to real influenza data from Japanese prefectures (July 2024–December 2025) with one-week-ahead forecasting simulations.

Significance. If the particle-filter methods can be shown to produce reliable joint state-parameter estimates from incomplete observations, the framework would supply a practical tool for real-time epidemic monitoring and short-term forecasting on spatially structured stochastic models, with direct relevance to public-health decision support.

major comments (2)
  1. [Abstract] Abstract: the central claim that the PF framework is 'effective' for real-time monitoring, forecasting, and adaptive decision-making is not accompanied by any quantitative error metrics, baseline comparisons, or goodness-of-fit statistics on either the synthetic or real-data experiments, leaving the effectiveness assertion without measurable support.
  2. [Abstract] Abstract / Methods (synthetic validation paragraph): the description of the two PF methodologies does not specify how the particle filter handles the joint estimation of transmission/recovery rates together with the latent state on the lattice, nor whether the reported performance degrades under the partial-observation regimes that are the paper’s stated target.
minor comments (1)
  1. [Abstract] Abstract: the data-collection interval 'July 2024 to December 2025' should be checked for typographical error or clarified as a projection, since the manuscript’s submission context appears to predate the end of this window.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify key aspects of our presentation. We provide point-by-point responses below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the PF framework is 'effective' for real-time monitoring, forecasting, and adaptive decision-making is not accompanied by any quantitative error metrics, baseline comparisons, or goodness-of-fit statistics on either the synthetic or real-data experiments, leaving the effectiveness assertion without measurable support.

    Authors: We agree that the abstract would benefit from explicit quantitative support. In the revised manuscript we will add concise references to error metrics (e.g., RMSE on state and parameter estimates from the synthetic experiments) and one-week-ahead forecast accuracy measures from the Japanese influenza application, together with a brief mention of baseline comparisons performed in the main text. revision: yes

  2. Referee: [Abstract] Abstract / Methods (synthetic validation paragraph): the description of the two PF methodologies does not specify how the particle filter handles the joint estimation of transmission/recovery rates together with the latent state on the lattice, nor whether the reported performance degrades under the partial-observation regimes that are the paper’s stated target.

    Authors: We accept that additional detail is warranted. The current methods section describes the two variants but does not explicitly state that parameters are jointly estimated by state augmentation within the particle filter. We will revise both the abstract and the synthetic-validation paragraph to clarify the augmented-state mechanism and to report performance separately for the partial-spatial-information regime, including any observed degradation relative to the infected-count variant. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper introduces a 2D lattice stochastic epidemic model and a particle filter framework for state/parameter estimation from partial observations. It validates the methods on synthetic data before applying the first method to external real-world influenza incidence data from Japanese prefectures (July 2024–December 2025) and performs one-week-ahead forecasts. No equations, ansatzes, or uniqueness claims are shown to reduce by construction to fitted parameters or self-citations; the central effectiveness claim rests on empirical performance against independent data sources. This matches the provided reader's assessment of score 2.0 with no load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Based on abstract only; the central claim rests on standard domain assumptions in stochastic epidemic modeling and the effectiveness of particle filters for this setting, with free parameters such as transmission rates estimated from data but not enumerated.

free parameters (1)
  • transmission and recovery rates
    Unknown parameters in the epidemic model that are estimated sequentially via the particle filter.
axioms (1)
  • domain assumption Disease spread follows a stochastic process on a two-dimensional lattice graph with partial observations.
    Core modeling premise invoked to justify the graph-based framework and the two PF methodologies.

pith-pipeline@v0.9.1-grok · 5703 in / 1272 out tokens · 31476 ms · 2026-06-28T23:53:59.188335+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

30 extracted references

  1. [1]

    Mathematical modeling for estimating influenza vaccine efficacy: A case study of the Valencian Community, Spain

    Andreu-Vilarroig, C., Villanueva, R. J. and González-Parra, G. “Mathematical modeling for estimating influenza vaccine efficacy: A case study of the Valencian Community, Spain.”In- fectious Disease Modelling, 9(3), 744–762, 2024

    2024

  2. [2]

    Mathematical modeling of influenza dynamics: Integrating seasonality and gradual waning immunity

    Andreu-Vilarroig, C., González-Parra, G. and Villanueva, R. J. “Mathematical modeling of influenza dynamics: Integrating seasonality and gradual waning immunity.”Bulletin of Math- ematical Biology, 87(6), 2025

    2025

  3. [3]

    A multi-model ensemble Kalman filter for data assimilation and fore- casting

    Bach, E. and Ghil, M. “A multi-model ensemble Kalman filter for data assimilation and fore- casting.”Journal of Advances in Modeling Earth Systems, 15(1), e2022MS003123, 2023

    2023

  4. [4]

    A network with tunable clustering, degree correlation and degree distribution, and an epidemic thereon

    Ball, F., Britton, T. and Sirl, D. “A network with tunable clustering, degree correlation and degree distribution, and an epidemic thereon.”Journal of Mathematical Biology, 66(4), 979– 1019, 2013

    2013

  5. [5]

    Network epidemi- ological analysis of COVID-19 transmission patterns by age, occupation and residence across four waves in Cyprus

    Dimitriou, P. A., Silvestros, V., Constantinou, E., Pitris, C. and Kolios, P. “Network epidemi- ological analysis of COVID-19 transmission patterns by age, occupation and residence across four waves in Cyprus.”Scientific Reports, 15(1), 28416, 2025

    2025

  6. [6]

    A tutorial on particle filtering and smoothing: Fifteen years later

    Doucet, A. and Johansen, A. M. “A tutorial on particle filtering and smoothing: Fifteen years later.”Handbook of Nonlinear Filtering, 12(1), 1–41, 2009

    2009

  7. [7]

    Data assimilation for agent-based models

    Ghorbani, A., Ghorbani, V., Nazari-Heris, M. and Asadi, S. “Data assimilation for agent-based models.”Mathematics, 11(20), 4296, 2023

    2023

  8. [8]

    Age-structured model of dengue transmission dynamics with time- varying parameters, and its application to Brazil

    Haq, I. U. and Richard, S. “Age-structured model of dengue transmission dynamics with time- varying parameters, and its application to Brazil.”Nonlinear Science, 100–133, 2026

    2026

  9. [9]

    Three basic epidemiological models

    Hethcote, H. W. “Three basic epidemiological models.”Applied Mathematical Ecology, 119–144, 1989

    1989

  10. [10]

    A stochastic agent-based model of the SARS-CoV-2 epidemic in France

    Hoertel, N., Blachier, M., Blanco, C., Olfson, M., Massetti, M., Sánchez Rico, M., Limosin, F. and Leleu, H. “A stochastic agent-based model of the SARS-CoV-2 epidemic in France.” Nature Medicine, 26(9), 1417–1421, 2020. 33

    2020

  11. [11]

    Particle filter-based data assimilation in dynamic data-driven simulation: Sensitivity analysis of three critical experimental conditions

    Huang, Y., Xie, X., Cho, Y. and Verbraeck, A. “Particle filter-based data assimilation in dynamic data-driven simulation: Sensitivity analysis of three critical experimental conditions.” Simulation, 99(4), 403–415, 2023

    2023

  12. [12]

    A minimal model for household effects in epidemics

    Huber, G., Kamb, M., Kawagoe, K., Li, L. M., Veytsman, B., Yllanes, D. and Zigmond, D. “A minimal model for household effects in epidemics.”Physical Biology, 17(6), 065010, 2020

    2020

  13. [13]

    Extinction and sta- tionary distribution of a stochastic COVID-19 epidemic model with time-delay

    Ikram, R., Khan, A., Zahri, M., Saeed, A., Yavuz, M. and Kumam, P. “Extinction and sta- tionary distribution of a stochastic COVID-19 epidemic model with time-delay.”Computers in Biology and Medicine, 141(1), 105–115, 2022

    2022

  14. [14]

    Assessing the impact of disease incidence and immunization on the resilience of complex networks during epidemics

    Islam, M. S., Ullah, M. S. and Kabir, K. A. “Assessing the impact of disease incidence and immunization on the resilience of complex networks during epidemics.”Infectious Disease Mod- elling, 10(1), 1–27, 2025

    2025

  15. [15]

    A comparative study of deterministic and stochastic computational modeling approaches for analyzing and optimizing COVID-19 control

    Kadri, A., Boudaoui, A., Ullah, S., Asiri, M., Saqib, A. B. and Riaz, M. B. “A comparative study of deterministic and stochastic computational modeling approaches for analyzing and optimizing COVID-19 control.”Scientific Reports, 15(1), 11710, 2025

    2025

  16. [16]

    Mathematical models of the spread of infection

    Kendall, D. G. “Mathematical models of the spread of infection.”Mathematics and Computer Science in Biology and Medicine, 213–225, 1965

    1965

  17. [17]

    Concurrency measures in the era of temporal network epidemiology: A review

    Masuda, N., Miller, J. C. and Holme, P. “Concurrency measures in the era of temporal network epidemiology: A review.”Journal of The Royal Society Interface, 18(179), 2021

    2021

  18. [18]

    Bridging compartmental models and network analysis in epidemiological modelling

    Mendes, J. M., Baptista, H. and MacNab, Y. C. “Bridging compartmental models and network analysis in epidemiological modelling.”Epidemiological Modelling, 2024

    2024

  19. [19]

    Stochastic epidemic models inference and diagnosis with Poisson random measure data augmentation

    Nguyen-Van-Yen, B., Del Moral, P. and Cazelles, B. “Stochastic epidemic models inference and diagnosis with Poisson random measure data augmentation.”Mathematical Biosciences, 335(1), 108583, 2021

    2021

  20. [20]

    An age-structured extension to the vectorial capacity model

    Novoseltsev, V. N., Michalski, A. I., Novoseltseva, J. A., Yashin, A. I., Carey, J. R. and Ellis, A. M. “An age-structured extension to the vectorial capacity model.”PLOS ONE, 7(6), e39479, 2012

    2012

  21. [21]

    Real-time growth rate for general stochastic SIR epidemics on unclustered networks

    Pellis, L., Spencer, S. E. F. and House, T. “Real-time growth rate for general stochastic SIR epidemics on unclustered networks.”Mathematical Biosciences, 265(1), 65–81, 2015. 34

    2015

  22. [22]

    Determining the rate of infectious disease testing through contagion potential

    Roy, S., Biswas, P. and Ghosh, P. “Determining the rate of infectious disease testing through contagion potential.”PLOS Global Public Health, 3(8), e0002229, 2023

    2023

  23. [23]

    Spatial-temporal dynamics in nonlocal epidemiological models

    Ruan, S. “Spatial-temporal dynamics in nonlocal epidemiological models.”Mathematics for Life Science and Medicine, 97–122, 2007

    2007

  24. [24]

    Data-assimilation and state estimation for contact-based spreading processes using the ensemble Kalman filter: Application to COVID- 19

    Schaum, A., Bernal-Jaquez, R. and Ramos, L. A. “Data-assimilation and state estimation for contact-based spreading processes using the ensemble Kalman filter: Application to COVID- 19.”Chaos, Solitons & Fractals, 157(1), 111887, 2022

    2022

  25. [25]

    Analysis of COVID-19 spread in Tokyo through an agent-based model with data assimilation

    Sun, C., Miyoshi, T., Richard, S. and Tsuzu, N. “Analysis of COVID-19 spread in Tokyo through an agent-based model with data assimilation.”J. Clin. Med.11, 2401, 2022

    2022

  26. [26]

    Data assim- ilation and agent-based modelling: Towards the incorporation of categorical agent parameters

    Ternes, P., Ward, J. A., Heppenstall, A., Kumar, V., Kieu, L. M. and Malleson, N. “Data assim- ilation and agent-based modelling: Towards the incorporation of categorical agent parameters.” Open Research Europe, 1(1), 131, 2022

    2022

  27. [27]

    Open problems in mathematical biology

    Vittadello, S. T. and Stumpf, M. P. H. “Open problems in mathematical biology.”Mathematical Biosciences, 354, 108926, 2022

    2022

  28. [28]

    Spatially structured models of viral dynamics: A scoping review

    Williams, T., McCaw, J. M. and Osborne, J. M. “Spatially structured models of viral dynamics: A scoping review.”Microbiology and Molecular Biology Reviews, 89(4), e00283–24, 2025

    2025

  29. [29]

    Computation of the basic reproduction numbers for reaction-diffusion epidemic models

    Yang, C. and Wang, J. “Computation of the basic reproduction numbers for reaction-diffusion epidemic models.”Mathematical Biosciences and Engineering, 20(8), 15201, 2023

    2023

  30. [30]

    InfectiousDiseasesWeeklyReport: Provisionaldata

    NationalInstituteofInfectiousDiseases.“InfectiousDiseasesWeeklyReport: Provisionaldata.” Available at:https://id-info.jihs.go.jp/surveillance/idwr/provisional/2025/47/ index.html. 35

    2025