Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Estimating Unobservable States in Stochastic Epidemic Models with Partial Information

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read An extended Kalman filter can nowcast the hidden infections in a stochastic Covid-19 model from reported data alone.

desk verdict Solid, narrowly scoped methods paper: useful dark-figure initialization and a clean EKF derivation, but the accuracy claim needs stronger numerical evidence than a single simulated path. read the letter →

arxiv 2506.00906 v1 pith:76NSBVPX submitted 2025-06-01 q-bio.PE math.PRmath.STstat.TH

Florent Ouabo Kamkumo , Ibrahim Mbouandi Njiasse , Ralf Wunderlich This is my paper
classification q-bio.PEmath.PRmath.STstat.TH MSC 92D3092-1060J6060G3562M20
keywords stochasticepidemicmodeldiffusionapproximationpartialinformationextendedKalmanfilterdarkfigureCovid-19nowcastinghiddenstateestimationcompartmental
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 argues that the hidden part of an epidemic, especially the number of undetected infected people $I^-$, can be estimated online from routinely reported data such as detected infections, hospitalizations, ICU occupancy, deaths, recoveries, and vaccinations. The setting is a stochastic Covid-19 model derived as a diffusion approximation of a continuous-time Markov chain, with the state vector split into observable and unobservable components. To handle the nonlinear drift, the authors linearize around the current filter estimate, freeze the noise coefficients at that estimate, and apply the conditional-Gaussian Kalman recursions. In simulations calibrated to German Covid-19 data, the filter begins with poor initial guesses but corrects itself within about two weeks and then tracks the undetected infections closely, with a narrowing confidence band. If this carries over to practice, health authorities could nowcast the 'dark figure' of unreported infections from data they already collect.

What carries the argument

The load-bearing object is a pair of stochastic recursions obtained by Euler-Maruyama discretization of a diffusion-approximated compartment model, split into a hidden component $Y$ and an observable component $Z$. Around this the paper builds the extended Kalman filter: at each time step it linearizes the quadratic infection drift $\beta y_1y_5/N$ around the current estimate $\hat M_n$, replaces $Y_n$ by $\hat M_n$ in the diffusion coefficients $\sigma, g, \ell$, and feeds the resulting conditionally Gaussian system into the Liptser-Shiryaev Kalman recursions, using the Moore-Penrose pseudoinverse to handle singular observation noise. Initialization uses dark-figure coefficients $U^1=I^-/I^+$ and $U^2=R^-_1/(R^+_1+R^+_2+R^+_3)$, modeled as conditionally Gaussian expert views, to set the initial hidden-state mean and a closed-form covariance matrix consistent with a constant total population size.

What would settle it

A decisive check would be a Monte Carlo study across many simulated outbreaks comparing the EKF's 95% confidence bands with the true hidden state; if the empirical coverage is far below 95% or the mean bias in $I^-$ exceeds the filter's reported standard deviation after the learning phase, the claim of high-accuracy tracking is refuted. A complementary real-data test is to apply the filter to a region whose undetected infections were later reconstructed from seroprevalence or exhaustive testing and compare the nowcast with that reconstruction.

Watch

Extended reading notes

Core claim

The central claim is that the extended Kalman filter applied to the diffusion-approximated extended Covid-19 model with cascade compartments produces accurate online estimates of the unobservable state vector, most importantly the number of undetected infected individuals $I^-$, after a short learning phase. The numerical evidence consists of simulated paths in which the filter estimate hugs the true signal while the initially large uncertainty, encoded by dark-figure coefficients with mean 10 and variance 25, shrinks within days. The paper also shows how the filter behaves under misspecified initial estimates: an overconfident expert with an inaccurate mean learns slowly, while a perfectly informed start is soon degraded by observation noise. Adding observable cascade compartments for post-recovery and post-vaccination immunity reduces estimation variance, with most of the reduction coming from the first cascade compartment.

Load-bearing premise

The load-bearing premise is that the linearization and frozen-noise approximation of the extended Kalman filter stay accurate for the Covid-19 model at the calibrated parameter values, so the filter's tracking of $I^-$ is not a numerical artifact of approximation error.

Editorial extensions

If this is right

  • If the filter's tracking accuracy holds, epidemic managers could estimate undetected infections in near real time from daily reported quantities, without relying on seroprevalence surveys or exhaustive testing.
  • An overconfident but wrong initial estimate delays convergence, so honest quantification of prior uncertainty is operationally relevant for nowcasting.
  • Adding observable cascade compartments for post-recovery and post-vaccination immunity reduces estimation variance, but the gain saturates quickly because transitions between cascade states are deterministic.
  • The methodology is generic for compartmental epidemic models whose observation drift is linear in the hidden state, since only the hidden-state drift needs to be linearized.
  • The filter's conditional covariance provides a real-time uncertainty band for the dark figure, which could inform testing intensity and intervention decisions.

Reading between the lines

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

  • Beyond the paper: the same construction could be applied to other respiratory diseases, but the learning time and bias would depend on how strongly the hidden state is coupled to observations, for instance through the testing rate.
  • Beyond the paper: the current evidence is entirely simulated, so a decisive real-world test would compare filter nowcasts of $I^-$ against later retrospective reconstructions from seroprevalence data or exhaustive testing in a specific region.
  • Beyond the paper: the filter's confidence band measures only filter uncertainty under the assumed model; parameter calibration error and model misspecification would likely make real-world intervals wider than those reported.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper develops a discrete-time diffusion-approximated stochastic epidemic model with partial information, decomposes the state into hidden and observable components, and proposes an extended Kalman filter (EKF) for estimating the hidden states. It extends a Covid-19 model with cascade compartments that represent the time spent in post-recovery and post-vaccination immunity, derives the EKF recursions by linearizing the quadratic infection drift and freezing the diffusion coefficients, and provides an initialization procedure based on dark figure coefficients. The numerical section presents simulated paths of the extended model and claims that the filter tracks the number of undetected infected individuals with high accuracy after a short learning phase.

Significance. If the numerical claim is substantiated, the paper offers a computationally light alternative to MCMC and particle filtering for nowcasting dark figures, and it usefully demonstrates how partially observable cascade inflows can be exploited by a filter. The algebraic derivation of the linearized recursion (Lemma 3.7) and the initialization covariance (Lemma 4.2) is clean, and the cascade-state idea is well motivated. However, the validation as it stands is an in-silico self-consistency check on a single simulated path, so the practical significance and the accuracy claim are not yet established at the level claimed in Section 5.

major comments (5)
  1. [§5.2] The central accuracy claim rests on a single simulated path generated from the same model that the filter presupposes. Figures 5.1–5.5 show one realization with no Monte Carlo repetitions, no random seed or code, and no quantitative error metric, so a favorable draw cannot be ruled out. Please add ensemble statistics (for example, RMSE or MAE over many independent paths), coverage rates of the reported bands, and preferably a comparison with a particle filter or bootstrap filter on the same observations.
  2. [§4.2, Lemma 4.2] The initialization formula is inconsistent with the definition of the second dark figure coefficient in Eq. (4.1). With Z=(I+,H,C,D,R+1,R+2,R+3,V1,V2,V3), the entry Z^2_0 is H, whereas U^2_0=R^-1,0/(R+1,0+R+2,0+R+3,0); therefore M^2_0 should be M^2_U(Z^5_0+Z^6_0+Z^7_0), not M^2_U Z^2_0, and the analogous correction is needed in the covariance matrix Q0. Table 5.2 suggests the implementation used the correct denominator, so this is likely a typo, but as written the lemma is not implementable without additional interpretation.
  3. [§3.3, Remark 3.8] The EKF approximation error is not controlled for the model and parameters at hand. Lemma 3.7 linearizes the quadratic drift and freezes sigma, g, and ell at the filter estimate; first-order accuracy is asserted by citing Picard [33] and the working paper [30], but no verification of the small-noise or regularity hypotheses is given for the parameter values in Table 5.1, and no bound on the Taylor remainder or on the frozen-diffusion error is supplied. The 95% band in Figure 5.2 is the conditional covariance of the approximate linear-Gaussian filter, not an error bound for the true posterior; please provide either a numerical comparison with a particle filter or a dedicated error analysis, or explicitly restrict the claims to the approximate filter.
  4. [§2.3 and §5.4] The cascade aggregation uses the uniform-age approximation psi_j=1/P_j, which the authors themselves state can lead to errors during rapid epidemic changes. Because the numerical ground truth is generated from the aggregated model, the reported tracking accuracy does not cover the model-approximation error, and the Covid-19 wave regime is exactly where both the model error and the EKF linearization error are least controlled. An experiment with the unaggregated cascade, with model mismatch, or with real observations is needed before the public-health wording in Section 1 and Section 5 is supported.
  5. [§5.1] The evaluation is a self-consistency check: the parameters are calibrated in the companion paper [25] to German data, but the simulated observations are generated from the same diffusion model that the EKF uses. This is a legitimate first step, but it does not validate the method on real partial observations. Please add a real-data nowcast or a misspecification experiment with different observation noise, or soften the claim that the method supports public health measures.
minor comments (5)
  1. [Appendix A] The abbreviation list contains the typo 'Extended Kalmnan filter'; it should read 'Extended Kalman filter'.
  2. [Table 5.1] The last row reads 'Q1_U = Q1_U'; this should presumably be 'Q1_U = Q2_U' for the variance of the second dark figure coefficient.
  3. [Theorem 3.3] The sentence 'the conditional distribution of Y_n given F^Z_n is is multivariate Gaussian' contains a duplicated 'is'.
  4. [§5.2] The band in Figure 5.2 is an approximate conditional band under a Gaussian assumption; calling it a '95% confidence band' without a coverage assessment over repeated simulations is potentially misleading and should be reworded.
  5. [References] Reference [30] is a working paper and is used as a central justification for EKF convergence in Remark 3.8; its availability and status should be clarified, or the argument should be made self-contained.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the EKF validation is a genuine in-silico experiment, not a fitted-input prediction.

full rationale

The derivation chain is self-contained with respect to its central claim. Hidden states are simulated from the nonlinear recursion (2.6) with parameters calibrated in the companion paper [25] to German RKI data, and the EKF (Algorithm 3.1) receives only the observable Z paths and the model coefficients. The comparison in Figs. 5.2-5.5 therefore has real falsifiability content: if the Taylor linearization of beta*Y1*Y5/N or the frozen-noise approximation in Lemma 3.7 were inaccurate on the calibrated paths, the filter would not track I- even in this in-silico setting; indeed Scenario 2 in Section 5.3 shows that a misspecified initial covariance delays convergence. The initialization of the filter (Lemma 4.2) uses prior dark-figure-coefficient distributions with mean 10, whereas the true initial DFCs are 4 and 1.6, so the 'learning phase' is not the filter rediscovering a fitted input. No equation in the paper reduces to its input by construction, and no fitted parameter is renamed as a prediction. The self-citations [25] and [30] are used for the model derivation, calibration, and EKF convergence, but the convergence statement in Remark 3.8 is jointly attributed to the external Picard [33], and the numerical experiments provide direct evidence for the central accuracy claim. Hence no significant circularity.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The central numerical results depend on a large set of externally calibrated or hand-chosen parameters and on the assumption that the discrete diffusion model and its EKF linearization faithfully represent the epidemic. The only genuinely new derivation, Lemma 4.2, is assumption-driven: its Gaussianity relies on the dark-figure-coefficient prior in Assumption 4.1. No new physical entities are introduced; cascade compartments are bookkeeping devices for age-structured vaccination and recovery.

free parameters (6)
  • Time-dependent rates beta_n, alpha_n, mu_n = not reported in this paper (calibrated to German RKI data in [25])
    Transmission, testing, and vaccination rates drive the simulated epidemic; their calibrated paths are taken from the companion paper and are essential to the numerical results.
  • Constant transition rates (gamma-, gamma+, gamma_H, gamma_C, eta+, eta-, delta, kappa, rho_1-, rho_2-, rho_V) = values in Table 5.1 (e.g., gamma-=gamma+=1/14, gamma_H=0.048, gamma_C=0.02, delta=0.03, kappa=0.05)
    These rates are chosen from cited literature or assumed by hand; the hidden-state trajectories and therefore the filter evaluation depend directly on them.
  • Immunity duration and cascade grouping (L_V=L_R=90, d_V=d_R=3, P_j=30) = 90, 3, 30
    These define the cascade compartments; the paper's key comparison of base vs extended models depends on this grouping.
  • Initial dark figure coefficient priors (M_U^1=M_U^2=10, Q_U^1=Q_U^2=25) = mean 10, variance 25
    Expert priors for the initial ratios of undetected to detected individuals; they determine M0, Q0 in Lemma 4.2 and drive the warm-up phase.
  • Initial compartment sizes (Table 5.2) = e.g., I-_0=300, R-_1,0=200, I+_0=75, R+_1,0=50
    Simulation initial conditions for both hidden and observable states; the 'true' path used for filter evaluation starts from these values.
  • Total population size N = 10^6
    Assumed constant population; the normalization property (Lemma 4.2, Eq. C.1) derives the initial susceptible estimate from N and the other compartments.
assumptions (7)
  • standard math Convergence of CTMC epidemic models to diffusion processes via functional CLT (Britton-Pardoux, Anderson-Kurtz, Ethier-Kurtz).
    Invoked in Section 2.1 to justify replacing counting processes by the SDE (2.3); the paper does not reprove the limit theorem.
  • standard math Euler-Maruyama discretization gives an accurate discrete-time approximation (2.5) of the diffusion.
    Used throughout to obtain the recursions (2.6); discretization error is not quantified.
  • standard math Liptser-Shiryaev conditional Gaussian filtering theorems (Theorems 3.3-3.4) apply to the linearized system.
    Basis for the EKF recursions in Algorithm 3.1; the paper cites [28] rather than proving them.
  • domain assumption First-order Taylor linearization plus frozen diffusion coefficients yields a small EKF error.
    Assumed in Section 3.3 and Remark 3.8; regularity and small-noise conditions of Picard [33] and [30] are not verified for the quadratic drift beta*y1*y5/N and state-dependent sigma, g, and l.
  • ad hoc to paper Assumption 4.1: initial dark figure coefficients are independent Gaussian with given mean and variance, and R-_2 and V- start empty.
    This prior is the basis for the Gaussian initialization in Lemma 4.2; it is an expert-choice assumption, not derived from data.
  • ad hoc to paper Uniform-age approximation psi_j=1/P_j for aggregated cascade compartments.
    Used in recursions (2.7)-(2.8); the paper itself notes it is accurate only when vaccination and recovery ages are uniformly distributed and may introduce errors during rapid epidemic changes (Section 2.3).
  • domain assumption Observable compartments Z are exactly and noiselessly observed.
    The observation equation in (2.6) has no measurement noise or reporting delay; the abstract and introduction motivate the dark-figure problem with reporting distortions, but these are not modeled in the observations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Estimating Unobservable States in Stochastic Epidemic Models with Partial Information." pith.science (2026). https://pith.science/paper/76NSBVPX

@misc{pith2026250600906,
  author       = {Pith},
  title        = {Pith review of: Estimating Unobservable States in Stochastic Epidemic Models with Partial Information},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/76NSBVPX}},
  note         = {Machine review of arXiv:2506.00906}
}
read the original abstract

This article investigates stochastic epidemic models with partial information and addresses the estimation of current values of not directly observable states. The latter is also called nowcasting and related to the so-called "dark figure" problem, which concerns, for example, the estimation of unknown numbers of asymptomatic and undetected infections. The study is based on Ouabo Kamkumo et al. (2025), which provides detailed information about stochastic multi-compartment epidemic models with partial information and various examples. Starting point is a description of the state dynamics by a system of nonlinear stochastic recursions resulting from a time-discretization of a diffusion approximation of the underlying counting processes. The state vector is decomposed into an observable and an unobservable component. The latter is estimated from the observations using the extended Kalman filter approach in order to take into account the nonlinearity of the state dynamics. Numerical simulations for a Covid-19 model with partial information are presented to verify the performance and accuracy of the estimation method.

Figures

Figures reproduced from arXiv: 2506.00906 by the authors.

Figure 2.1
Figure 2.1. Base Covid-19 model with partial information consisting of of three fully hidden states I −,R −,S, two partially hidden states R +,V, and four observable states I +,H,C,D. k Transition Transition vectors ξ ⊤ k Intensity λi(t,X) 1 Infection of susceptible (1,0,0,0,−1,0,0,0,0) βS I − N = βY 5 Y 1 N 2 Test of infected undetected (−1,0,0,0,0,1,0,0,0) αI − = αY 1 3 Recovering of infected detected (0,0,1,0,0,−1,0,0,0) γ +… view at source ↗
Figure 2.2
Figure 2.2. Extended Covid-19 model with partial information consisting of the 5 hidden states I −,R − 1 ,R − 2 ,V −,S and d R + d V + 4 observable states. The four observable states I +,H,C,D were already present in the base Covid-19 model depicted in [PITH_FULL_IMAGE:figures/full_fig_p011_2_2.png] view at source ↗
Figure 5.1
Figure 5.1. Simulated paths of hidden and observable states of the extended Covid-19 model. the quantile of order 0.975 of the standard normal distribution. It can be seen that the filter estimate closely tracks the true signal, highlighting the accuracy and reliability of the EKF [PITH_FULL_IMAGE:figures/full_fig_p022_5_1.png] view at source ↗
Figures from the paper (6 more)
Figure 5.2
Figure 5.2. Figure 5.2: True hidden state I −, filter estimate Me1 = MeI − , and associated 95% confidence band showing that large initial uncertainty is reduced by learning from observations. in capturing the underlying epidemic dynamics. This figure reveals that the initially rather large…
Figure 5.3
Figure 5.3. Figure 5.3: Conditional standard deviations of the filter estimates of the five hidden states [PITH_FULL_IMAGE:figures/full_fig_p023_5_3.png]
Figure 5.4
Figure 5.4. Figure 5.4: Comparison of the effect of initial uncertainty on filtering performance. The figure shows for the three scenarios the true hidden state I −, the filter estimate Me1 = MeI − , and the associated 95% confidence band. Scenario 1: Poorly informed and uncertain expert. T…
Figure 5
Figure 5. Figure 5: a shows what was already observed in Figure 5.2 for the three-year period, [PITH_FULL_IMAGE:figures/full_fig_p025_5.png]
Figure 5.5
Figure 5.5. Figure 5.5: Standard deviation evolution over time To ensure consistency, the same initial values are used across models for the hidden compartments S,I −, and R − 1 respectively R −. For the observable recovered cascade com￾partments in the extended models, the initial value of…
Figure 5
Figure 5. Figure 5: a illustrates the evolution of the standard deviation over time for each model [PITH_FULL_IMAGE:figures/full_fig_p026_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Identifiability and Information-Based Inference for Epidemic Transmission Models Under Partial Observation

    math.ST 2026-07 conditional novelty 3.0 of 10

    Partial observation regimes determine which epidemic-network parameters are identifiable, and a missing-information decomposition quantifies the loss, with transmission and external infection confounded when exposure ...

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [33]

    Efficiency of the extended Kalman filter for nonlinear systems with small noise.SIAM Journal on applied mathematics 51, 3 (1991), 843–885

    PICARD, J. Efficiency of the extended Kalman filter for nonlinear systems with small noise.SIAM Journal on applied mathematics 51, 3 (1991), 843–885. 32 F. Ouabo Kamkumo, I. Mbouandi Njiasse, R. Wunderlich

    work page 1991

  2. [30]

    M., KAMKUMO, F

    NJIASSE, I. M., KAMKUMO, F. O.,ANDWUNDERLICH, R. Convergence analysis for the extended Kalman filter in continuous time.Working Paper(2025)

    work page 2025

  3. [25]

    Stochastic Epidemic Models with Partial Information

    KAMKUMO, F. O., NJIASSE, I. M.,ANDWUNDERLICH, R. Stochastic epidemic models with partial information.Preprint, arXiv:2503.07251(2025)

    work page Pith review arXiv 2025

  4. [1]

    Extended skew Kalman filters for COVID-19 pandemic state estimation

    ALYAMI, L.,ANDDAS, S. Extended skew Kalman filters for COVID-19 pandemic state estimation. In2023 Sixth International Conference of Women in Data Science at Prince Sultan University (WiDS PSU)(2023), Institute of Electrical and Electronics Engineers (IEEE)

    work page 2023

  5. [2]

    F.,ANDKURTZ, T

    ANDERSON, D. F.,ANDKURTZ, T. G. Continuous time Markov chain models for chemical reaction networks. InDesign and analysis of biomolecular circuits. Springer, 2011, pp. 3–42

    work page 2011

  6. [3]

    Tracking R of COVID-19: A new real-time estimation using the Kalman filter.PloS one 16, 1 (2021), e0244474

    ARROYO-MARIOLI, F., BULLANO, F., KUCINSKAS, S.,ANDRONDÓN-MORENO, C. Tracking R of COVID-19: A new real-time estimation using the Kalman filter.PloS one 16, 1 (2021), e0244474

    work page 2021

  7. [4]

    BAIN, A.,ANDCRISAN, D.Fundamentals of stochastic filtering, vol. 3. Springer, 2009

    work page 2009

  8. [5]

    N.Generalized inverses: theory and applications

    BEN-ISRAEL, A.,ANDGREVILLE, T. N.Generalized inverses: theory and applications. Springer Science & Business Media, 2006

    work page 2006

Show all 41 references
  1. [6]

    N., DEY, N.,ANDSANTOSH, K

    BHAPKAR, H., MAHALLE, P. N., DEY, N.,ANDSANTOSH, K. Revisited COVID-19 mortality and recovery rates: are we missing recovery time period?Journal of Medical Systems 44, 12 (2020), 202

    2020

  2. [7]

    Stochastic epidemic models: a survey.Mathematical Biosciences 225, 1 (2010), 24–35

    BRITTON, T. Stochastic epidemic models: a survey.Mathematical Biosciences 225, 1 (2010), 24–35

    2010

  3. [8]

    BRITTON, T.,ANDO’NEILL, P. D. Bayesian inference for stochastic epidemics in populations with random social structure. Scandinavian Journal of Statistics 29, 3 (2002), 375–390

    2002

  4. [9]

    Springer Internat

    BRITTON, T.,ANDPARDOUX, E., Eds.Stochastic Epidemic Models with Inference. Springer Internat. Publ., 2019

    2019

  5. [10]

    L., PARDOUX, E., SZNITMAN, A.-S.,ANDPARDOUX, E

    BURKHOLDER, D. L., PARDOUX, E., SZNITMAN, A.-S.,ANDPARDOUX, E. Filtrage non linéaire et équations aux dérivées partielles stochastiques associées. InEcole d’Eté de Probabilités de Saint-Flour XIX—1989(1991), Springer, pp. 68–163

    1991

  6. [11]

    Bayesian particle filter algorithm for learning epidemic dynamics

    CALVETTI, D., HOOVER, A., ROSE, J.,ANDSOMERSALO, E. Bayesian particle filter algorithm for learning epidemic dynamics. Inverse Problems 37, 11 (2021), 115008

    2021

  7. [12]

    Adaptive dynamic modelling of HIV/AIDS epidemic using extended Kalman filter.Journal of Biological Systems 3, 03 (1995), 759–768

    CAZELLES, B.,ANDCHAU, N. Adaptive dynamic modelling of HIV/AIDS epidemic using extended Kalman filter.Journal of Biological Systems 3, 03 (1995), 759–768

    1995

  8. [13]

    CHARPENTIER, A., ELIE, R., LAURIÈRE, M.,ANDTRAN, V. C. COVID-19 pandemic control: balancing detection policy and lockdown intervention under ICU sustainability.Preprint, arXiv:2005.06526(2020)

    2020 arXiv

  9. [14]

    On Kalman filtering for conditionally Gaussian systems with random matrices

    CHEN, H.-F., KUMAR, P.,ANDVANSCHUPPEN, J. On Kalman filtering for conditionally Gaussian systems with random matrices. Systems & Control Letters 13, 5 (1989), 397–404

    1989

  10. [15]

    CHEN, N., LI, Y.,ANDLIU, H. Conditional Gaussian nonlinear system: A fast preconditioner and a cheap surrogate model for complex nonlinear systems.Chaos: An Interdisciplinary Journal of Nonlinear Science 32, 5 (2022)

    2022

  11. [16]

    CHEN, N.,ANDMAJDA, A. J. Conditional Gaussian systems for multiscale nonlinear stochastic systems: Prediction, state esti- mation and uncertainty quantification.Entropy 20, 7 (2018), 509

    2018

  12. [17]

    Invisible infections: A partial information approach for estimating the transmission dynamics of the Covid-19 pandemic.arXiv preprint arXiv:2212.13443(2022)

    COLANERI, K., DAMIAN, C.,ANDFREY, R. Invisible infections: A partial information approach for estimating the transmission dynamics of the Covid-19 pandemic.arXiv preprint arXiv:2212.13443(2022)

    2022 arXiv

  13. [18]

    N.,ANDKURTZ, T

    ETHIER, S. N.,ANDKURTZ, T. G.Markov processes: characterization and convergence. John Wiley & Sons, 2009

    2009

  14. [19]

    M., LAYDON, D., NEDJATI-GILANI, G., IMAI, N., AINSLIE, K., BAGUELIN, M., BHATIA, S., BOONYASIRI, A., CUCUNUBÁ, Z., CUOMO-DANNENBURG, G.,ET AL

    FERGUSON, N. M., LAYDON, D., NEDJATI-GILANI, G., IMAI, N., AINSLIE, K., BAGUELIN, M., BHATIA, S., BOONYASIRI, A., CUCUNUBÁ, Z., CUOMO-DANNENBURG, G.,ET AL. Impact of non-pharmaceutical interventions (npis) to reduce COVID- 19 mortality and healthcare demand. imperial college C...

    2020

  15. [20]

    MIT press, 1974

    GELB, A.Applied optimal estimation. MIT press, 1974

    1974

  16. [21]

    Approximation of epidemic models by diffusion processes and their statistical inference

    GUY, R., LARÉDO, C.,ANDVERGU, E. Approximation of epidemic models by diffusion processes and their statistical inference. Journal of Mathematical Biology 70, 3 (2015), 621–646

    2015

  17. [22]

    Approximation and inference of epidemic dynamics by diffusion processes.Journal de la Société Française de Statistique 157, 1 (2016), 71–100

    GUY, R., LARÉDO, C.,ANDVERGU, E. Approximation and inference of epidemic dynamics by diffusion processes.Journal de la Société Française de Statistique 157, 1 (2016), 71–100

    2016

  18. [23]

    A new estimation method for COVID-19 time-varying reproduction number using active cases.Scientific Reports 12, 1 (2022), 6675

    HASAN, A., SUSANTO, H., TJAHJONO, V., KUSDIANTARA, R., PUTRI, E., NURAINI, N.,ANDHADISOEMARTO, P. A new estimation method for COVID-19 time-varying reproduction number using active cases.Scientific Reports 12, 1 (2022), 6675

    2022

  19. [24]

    L., BHADRA, A., ATCHADÉ, Y.,ANDKING, A

    IONIDES, E. L., BHADRA, A., ATCHADÉ, Y.,ANDKING, A. Iterated filtering.The Annals of Statistics(2011), 1776–1802

    2011

  20. [26]

    An application of the ensemble Kalman filter in epidemiological modelling.Plos one 16, 8 (2021), e0256227

    LAL, R., HUANG, W.,ANDLI, Z. An application of the ensemble Kalman filter in epidemiological modelling.Plos one 16, 8 (2021), e0256227

    2021

  21. [27]

    E.,ANDFINKENSTÄDT, B

    LEKONE, P. E.,ANDFINKENSTÄDT, B. F. Statistical inference in a stochastic epidemic SEIR model with control intervention: Ebola as a case study.Biometrics 62, 4 (2006), 1170–1177

    2006

  22. [28]

    S.,ANDSHIRYAEV, A

    LIPTSER, R. S.,ANDSHIRYAEV, A. N.Statistics of Random Processes II: Applications. Applications of Mathematics Stochastic Modelling and Applied Probability Series. Springer, 2001

    2001

  23. [29]

    S., HAARIO, H.,ANDTCHUENCHE, J

    NDANGUZA, D., MBALAWATA, I. S., HAARIO, H.,ANDTCHUENCHE, J. M. Analysis of bias in an Ebola epidemic model by extended Kalman filter approach.Mathematics and Computers in Simulation 142(2017), 113–129

    2017

  24. [31]

    M., KAMKUMO, F

    NJIASSE, I. M., KAMKUMO, F. O.,ANDWUNDERLICH, R. Stochastic optimal control of an epidemic under partial information. Preprint, arXiv:2503.06804(2025)

    2025 arXiv

  25. [32]

    D.,ANDROBERTS, G

    O’NEILL, P. D.,ANDROBERTS, G. O. Bayesian inference for partially observed stochastic epidemics.Journal of the Royal Statistical Society Series A: Statistics in Society 162, 1 (1999), 121–129

    1999

  26. [34]

    Estimation of the quadratic variation of nearly observed semimartingales with application to filtering.SIAM journal on control and optimization 31, 2 (1993), 494–517

    PICARD, J. Estimation of the quadratic variation of nearly observed semimartingales with application to filtering.SIAM journal on control and optimization 31, 2 (1993), 494–517

    1993

  27. [35]

    Community mitigation guidelines to prevent pandemic influenza—united states, 2017.MMWR

    QUALLS, N. Community mitigation guidelines to prevent pandemic influenza—united states, 2017.MMWR. Recommendations and reports 66(2017)

    2017

  28. [36]

    EKF-SIRD model algorithm for predicting the Coronavirus (COVID-19) spreading dynamics

    SEBBAGH, A.,ANDKECHIDA, S. EKF-SIRD model algorithm for predicting the Coronavirus (COVID-19) spreading dynamics. Scientific Reports 12, 1 (2022), 13415

    2022

  29. [37]

    Maximum likelihood-based extended Kalman filter for COVID-19 prediction.Chaos, Solitons & Fractals 146(2021), 110922

    SONG, J., XIE, H., GAO, B., ZHONG, Y., GU, C.,ANDCHOI, K.-S. Maximum likelihood-based extended Kalman filter for COVID-19 prediction.Chaos, Solitons & Fractals 146(2021), 110922

    2021

  30. [38]

    STREFTARIS, G.,ANDGIBSON, G. J. Bayesian inference for stochastic epidemics in closed populations.Statistical Modelling 4, 1 (2004), 63–75

    2004

  31. [39]

    Optimal control strategies based on extended Kalman filter in mathematical models of COVID-19.International Journal of Electrical & Computer Engineering (2088-8708) 14, 6 (2024)

    SUHIKA, D., SARAGIH, R., HANDAYANI, D.,ANDAPRI, M. Optimal control strategies based on extended Kalman filter in mathematical models of COVID-19.International Journal of Electrical & Computer Engineering (2088-8708) 14, 6 (2024)

    2024

  32. [40]

    Dynamics identification and forecasting of COVID-19 by switching Kalman filters.Computational Mechanics 66(2020), 1179–1193

    ZENG, X.,ANDGHANEM, R. Dynamics identification and forecasting of COVID-19 by switching Kalman filters.Computational Mechanics 66(2020), 1179–1193

    2020

  33. [41]

    Extended Kalman filter based on stochastic epidemiological model for COVID-19 modelling.Computers in Biology and Medicine 137(2021), 104810

    ZHU, X., GAO, B., ZHONG, Y., GU, C.,ANDCHOI, K.-S. Extended Kalman filter based on stochastic epidemiological model for COVID-19 modelling.Computers in Biology and Medicine 137(2021), 104810

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.