pith. sign in

arxiv: 2606.25902 · v2 · pith:BWNW2NXQnew · submitted 2026-06-24 · 📊 stat.AP

Count data modeling and forecasting of malaria incidence using generalized time series regression

Adithya B. Somaraj , Praveen D. Chougale , Usha Ananthakumar , Karthika M. Satyanarayanan This is my paper

Pith reviewed 2026-07-01 06:55 UTC · model grok-4.3

classification 📊 stat.AP
keywords malaria incidencecount dataGLARMAnegative binomialtime series forecastingoverdispersionserial dependenceurban surveillance
0
0 comments X

The pith

GLARMA negative binomial regression yields more stable and accurate forecasts of monthly malaria incidence than Poisson regression or standard time series methods.

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

The study models monthly Plasmodium vivax case counts from Mumbai HMIS data 2012-2019 together with meteorological covariates. Standard Poisson regression reveals strong environmental associations but fails due to overdispersion; negative binomial regression improves the fit yet still leaves residual serial correlation. Introducing the GLARMA structure to capture autoregressive and moving-average effects in the linear predictor produces forecasts that are evaluated by rolling time-series cross-validation and show consistently higher accuracy and lower variability than competing approaches. This joint treatment of overdispersion and temporal dependence matters for building early-warning systems that support timely public-health decisions in urban settings where malaria remains endemic.

Core claim

The GLARMA Negative binomial model consistently demonstrated superior predictive performance and greater predictive stability than competing regression and time series approaches.

What carries the argument

The Generalized Linear Autoregressive Moving Average (GLARMA) framework applied inside a negative binomial regression, which explicitly models serial dependence in the count series after accounting for overdispersion.

If this is right

  • Forecasts that jointly handle overdispersion and serial correlation support more reliable resource allocation for urban malaria control.
  • Rolling time-series cross-validation provides a practical way to compare count-data models for surveillance applications.
  • Seasonal effects emerge as stronger predictors than individual climatic variables once temporal dependence is modeled.
  • Simulation-based forecasting from the fitted GLARMA model supplies interval estimates that reflect both sources of variability.

Where Pith is reading between the lines

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

  • Similar GLARMA negative binomial specifications may improve short-term forecasting for other overdispersed infectious-disease count series that exhibit seasonal and serial patterns.
  • If reporting systems or control interventions change substantially, the identified covariate relationships would require re-estimation before continued use in operational forecasts.

Load-bearing premise

The 2012-2019 Mumbai surveillance records and meteorological series contain a stable relationship between incidence and covariates that will continue to hold for future periods without major shifts in reporting practices, interventions, or climate.

What would settle it

On post-2019 Mumbai malaria counts, the GLARMA negative binomial model no longer produces lower forecast error or higher stability than the Poisson, negative binomial, or ARIMA benchmarks under the same rolling cross-validation protocol.

Figures

Figures reproduced from arXiv: 2606.25902 by Adithya B. Somaraj, Karthika M. Satyanarayanan, Praveen D. Chougale, Usha Ananthakumar.

Figure 1
Figure 1. Figure 1: Baseline Poisson GLM residual diagnostics. [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Baseline negative binomial GLM residual diagnostics. [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Fig 3. In-sample actual versus predicted [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 3
Figure 3. Figure 3: In-sample actual versus predicted Plasmodium vivax cases. (Left) Predictions from the baseline Poisson GLM. (Right) Predictions from the baseline negative binomial GLM. Although the baseline NB model correctly captured the heteroscedasticity, it left short-term month-to-month momentum unaddressed. As shown in [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Temporal autocorrelation analysis of the randomized quantile [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Fig 5. Accuracy degradation across forecast horizons. [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Fig 6. Forecast stability across cross-validation folds. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 6
Figure 6. Figure 6: Forecast stability across cross-validation folds. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
read the original abstract

Malaria remains a major public health concern in many urban regions of India, where timely prediction of malaria incidence is essential for effective surveillance and resource allocation. This study examines count data approaches for understanding and predicting malaria incidence in the Mumbai region. The analysis used monthly $\textit{Plasmodium vivax}$ surveillance data from the Health Management Information System (HMIS) collected between 2012 and 2019, together with meteorological variables. Initial Poisson regression models suggested strong associations between malaria incidence and environmental factors; however, diagnostic assessment revealed substantial overdispersion, indicating that the Poisson model did not adequately capture the data's variability. Negative binomial regression provided a better representation of the data and indicated that seasonal effects were more strongly associated with malaria incidence than individual climatic covariates. Residual analyses further identified significant serial dependence not captured by baseline regression models. To address this limitation, a Generalized Linear Autoregressive Moving Average (GLARMA) framework was implemented to model temporal correlation explicitly. Forecasts were generated using simulation-based methods and evaluated through rolling time series cross-validation. The GLARMA Negative binomial model consistently demonstrated superior predictive performance and greater predictive stability than competing regression and time series approaches. These findings highlight the importance of jointly accounting for overdispersion and serial dependence in malaria surveillance data and demonstrate the value of count time series models for supporting early warning systems in urban settings.

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

0 major / 4 minor

Summary. The manuscript analyzes monthly Plasmodium vivax malaria incidence counts from Mumbai HMIS (2012-2019) together with meteorological covariates. It begins with Poisson regression, identifies overdispersion, moves to negative binomial regression, detects residual serial dependence, introduces a GLARMA negative binomial specification to capture both features, generates simulation-based forecasts, and evaluates them via rolling time-series cross-validation. The central claim is that the GLARMA negative binomial model exhibits superior predictive performance and stability relative to competing regression and time-series approaches.

Significance. If the reported superiority is reproducible, the work supplies a concrete, field-appropriate demonstration that jointly modeling overdispersion and autoregressive-moving-average dependence improves short-term count forecasts for urban malaria surveillance. Such an application is directly relevant to early-warning systems and resource allocation in endemic settings.

minor comments (4)
  1. The abstract asserts superior performance of the GLARMA negative binomial model but does not report any numerical values (e.g., mean absolute error, coverage rates, or interval scores) from the rolling cross-validation; these should be added to the abstract or a dedicated results paragraph for immediate readability.
  2. Section describing the GLARMA specification should include the exact order (p,q) selected, the link function, and whether the dispersion parameter is estimated jointly or profiled; the current description leaves these choices implicit.
  3. The rolling cross-validation procedure is mentioned but lacks explicit details on window length, forecast horizon, and how the simulation-based predictive distributions are scored; a short algorithmic box or table would clarify reproducibility.
  4. Figure captions for residual diagnostics and forecast plots should state the exact sample sizes and periods used so readers can map them to the 2012-2019 data range without ambiguity.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the supportive summary of our work and the recommendation of minor revision. No specific major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper's central claim rests on empirical model comparison via rolling time-series cross-validation applied to the 2012-2019 HMIS data after sequential diagnostics (Poisson overdispersion check → negative binomial → GLARMA for residuals). This pipeline is a standard, self-contained statistical workflow whose superiority result on held-out folds does not reduce by construction to the fitted inputs or any self-citation chain. No equations or steps match the enumerated circularity patterns; the evaluation uses out-of-sample folds rather than in-sample refits presented as predictions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides insufficient detail to enumerate specific free parameters, axioms, or invented entities beyond the standard statistical assumptions of generalized linear models and time series dependence.

pith-pipeline@v0.9.1-grok · 5794 in / 1054 out tokens · 30367 ms · 2026-07-01T06:55:48.463337+00:00 · methodology

Review history (2 revisions) →

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

28 extracted references · 15 canonical work pages

  1. [1]

    Malaria early warnings based on seasonal climate forecasts from multi-model ensembles

    Thomson MC, Doblas-Reyes FJ, Mason SJ, et al. Malaria early warnings based on seasonal climate forecasts from multi-model ensembles. Nature. 2006;439(7076):576-9. doi:10.1038/nature04503

    work page doi:10.1038/nature04503 2006

  2. [2]

    World malaria report 2024

    World Health Organization. World malaria report 2024. World Health Organization. 2024

    2024

  3. [3]

    Regression models for count data in R

    Zeileis A, Kleiber C, Jackman S. Regression models for count data in R. Journal of Statistical Software. 2008;27(8):1-25. doi:10.18637/jss.v027.i08

    work page doi:10.18637/jss.v027.i08 2008

  4. [4]

    Mapping the global endemicity and clinical burden ofPlasmodium vivax, 2000–17: a spatial and temporal modelling study

    Battle KE, Lucas TC, Nguyen M, et al. Mapping the global endemicity and clinical burden ofPlasmodium vivax, 2000–17: a spatial and temporal modelling study. The Lancet. 2019;394(10195):332-43. doi:10.1016/S0140-6736(19)31096-7

    work page doi:10.1016/s0140-6736(19)31096-7 2000

  5. [5]

    Determinants of relapse periodicity inPlasmodium vivaxmalaria

    White NJ. Determinants of relapse periodicity inPlasmodium vivaxmalaria. Malaria Journal. 2011;10:297. doi:10.1186/1475-2875-10-297

    work page doi:10.1186/1475-2875-10-297 2011

  6. [6]

    Malaria situation in India with special reference to tribal areas

    Sharma VP. Malaria situation in India with special reference to tribal areas. Indian Journal of Medical Research. 2009;130(2):98-106. July 1, 2026 14/16

    2009

  7. [7]

    Generalized Linear Models

    Nelder JA, Wedderburn RWM. Generalized Linear Models. Journal of the Royal Statistical Society: Series A (General). 1972;135(3):370-84

    1972

  8. [8]

    Negative Binomial Regression

    Hilbe JM. Negative Binomial Regression. 2nd ed. Cambridge: Cambridge University Press; 2011

    2011

  9. [9]

    Predicting Malaria Cases in Mumbai: Insights from Statistical and Machine Learning Models

    Chougale PD, Somaraj AB, Ananthakumar U. Predicting Malaria Cases in Mumbai: Insights from Statistical and Machine Learning Models. In: Pattern Recognition. ICPR 2024 International Workshops and Challenges. vol. 15615 of Lecture Notes in Computer Science. Cham: Springer; 2025. p. 19-34. doi:10.1007/978-3-031-87660-8 2

    work page doi:10.1007/978-3-031-87660-8 2024

  10. [10]

    Effects of short- and long-term climatic variability on malaria transmission dynamics

    Bri¨ et OJT, Vounatsou P, Gunawardena DM, Galappaththy GNL, Amerasinghe PH. Effects of short- and long-term climatic variability on malaria transmission dynamics. European Journal of Epidemiology. 2008;23:557-66. doi:10.1007/s10654-008-9271-y

    work page doi:10.1007/s10654-008-9271-y 2008

  11. [11]

    Regression Models for Time Series Analysis

    Kedem B, Fokianos K. Regression Models for Time Series Analysis. Hoboken, NJ: John Wiley & Sons; 2002

    2002

  12. [12]

    Time Series Models Based on Generalized Linear Models: Some Further Results

    Li WK. Time Series Models Based on Generalized Linear Models: Some Further Results. Biometrics. 1994;50(2):506-11. doi:10.2307/2533393

    work page doi:10.2307/2533393 1994

  13. [13]

    Observation-driven models for Poisson counts

    Davis RA, Dunsmuir WTM, Streett SB. Observation-driven models for Poisson counts. Biometrika. 2003;90(4):777-90. doi:10.1093/biomet/90.4.777

    work page doi:10.1093/biomet/90.4.777 2003

  14. [14]

    Theglarmapackage for observation-driven time series regression of counts

    Dunsmuir WTM, Scott DJ. Theglarmapackage for observation-driven time series regression of counts. Journal of Statistical Software. 2015;67(7):1-36. doi:10.18637/jss.v067.i07

    work page doi:10.18637/jss.v067.i07 2015

  15. [15]

    Robust estimate for count time series using GLARMA models: An application to environmental and epidemiological data

    Camara AJA, Reisen VA, Bondon P. Robust estimate for count time series using GLARMA models: An application to environmental and epidemiological data. Applied Mathematical Modelling. 2025;137:115658. doi:10.1016/j.apm.2024.115658

    work page doi:10.1016/j.apm.2024.115658 2025

  16. [16]

    Combining generalized linear autoregressive moving average and bootstrap models for analyzing time series of respiratory diseases and air pollutants

    Camara AJA, Reisen VA, Franco GC, Bondon P. Combining generalized linear autoregressive moving average and bootstrap models for analyzing time series of respiratory diseases and air pollutants. Mathematics. 2025;13(5):859. doi:10.3390/math13050859

    work page doi:10.3390/math13050859 2025

  17. [17]

    Modeling and Forecasting Indian Malaria Incidence Using Generalized Time Series Models

    Mukhopadhyay S, Tiwari R, Shetty P, Gogtay NJ, Thatte UM. Modeling and Forecasting Indian Malaria Incidence Using Generalized Time Series Models. Communications in Statistics: Case Studies, Data Analysis and Applications. 2019;5(2):111-20. doi:10.1080/23737484.2019.1580629

    work page doi:10.1080/23737484.2019.1580629 2019

  18. [18]

    Count Data in Time Series [Bachelor thesis]; 2010

    Bel K. Count Data in Time Series [Bachelor thesis]; 2010. Available from: https://thesis.eur.nl/pub/7507/7507-Bel.pdf

    2010

  19. [19]

    Forecasting Malaria in Indian States: A Time Series Approach with R Shiny Integration

    Ghosh SK, Ananthakumar U, Chougale PD, Somaraj AB. Forecasting Malaria in Indian States: A Time Series Approach with R Shiny Integration. Malaria Journal. 2025;24:403. doi:10.1186/s12936-025-05526-z

    work page doi:10.1186/s12936-025-05526-z 2025

  20. [20]

    Time Series Analysis: Forecasting and Control

    Box GEP, Jenkins GM, Reinsel GC, Ljung GM. Time Series Analysis: Forecasting and Control. 5th ed. Hoboken, New Jersey: John Wiley & Sons; 2015

    2015

  21. [21]

    Analysis of Financial Time Series

    Tsay RS. Analysis of Financial Time Series. 3rd ed. Hoboken, New Jersey: John Wiley & Sons; 2010. July 1, 2026 15/16

    2010

  22. [22]

    Randomized quantile residuals

    Dunn PK, Smyth GK. Randomized quantile residuals. Journal of Computational and Graphical Statistics. 1996;5(3):236-44. doi:10.2307/1390802

    work page doi:10.2307/1390802 1996

  23. [23]

    Empirical Investigation of Randomized Quantile Residuals for Diagnosis of Non-Normal Regression Models [Master’s thesis]

    Sadeghpour A. Empirical Investigation of Randomized Quantile Residuals for Diagnosis of Non-Normal Regression Models [Master’s thesis]. Saskatoon, Saskatchewan, Canada; 2016. Available from: https://harvest.usask.ca/handle/10388/7404

    2016

  24. [24]

    Evaluating Density Forecasts with Applications to Financial Risk Management

    Diebold FX, Gunther TA, Tay AS. Evaluating Density Forecasts with Applications to Financial Risk Management. International Economic Review. 1998;39:863-83. doi:10.2307/2527342

    work page doi:10.2307/2527342 1998

  25. [25]

    DHARMa: residual diagnostics for hierarchical (multi-level/mixed) regression models; 2022

    Hartig F. DHARMa: residual diagnostics for hierarchical (multi-level/mixed) regression models; 2022. R package version 0.4.6. Available from: https://CRAN.R-project.org/package=DHARMa

    2022

  26. [26]

    Forecasting: Principles and Practice

    Hyndman RJ, Athanasopoulos G. Forecasting: Principles and Practice. 3rd ed. Melbourne, Australia: OTexts; 2021. Available from: https://otexts.com/fpp3/

    2021

  27. [27]

    Health Management Information System (HMIS) Portal

    Ministry of Health and Family Welfare. Health Management Information System (HMIS) Portal;. Government of India. Available from: https://hmis.mohfw.gov.in/

  28. [28]

    Projection of district-level annual population by quinquennial age-group and sex from 2012 to 2031 in India

    Dhar M. Projection of district-level annual population by quinquennial age-group and sex from 2012 to 2031 in India. Mumbai, India: International Institute for Population Sciences (IIPS); 2022. Available from:https://www.iipsindia.ac. in/sites/default/files/FULL_REPORT_WITH_FINAL_TABLES.pdf. July 1, 2026 16/16

    2012