Pith. sign in

REVIEW 4 major objections 5 minor 112 references

This paper claims that spatio-temporal series with normal, Poisson, binomial, gamma, or inverse-Gaussian marginals can be modeled by two coupled GLM-type equations—one for the conditional mean and one for the dispersion—and that joint quasi

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 07:52 UTC pith:IBNIDUP2

load-bearing objection Worth engaging with, but the paper's own simulations show the count-dispersion inference is unreliable at low counts, and the asymptotics are conjectured. the 4 major comments →

arxiv 2607.08276 v2 pith:IBNIDUP2 submitted 2026-07-09 stat.CO stat.ME

glmSTARMA -- An R-Package for fitting autoregressive spatio-temporal models following generalized linear models

Steffen Maletz , Konstantinos Fokianos , Roland Fried This is my paper
classification stat.CO stat.ME MSC 62M1062M3062-04
keywords spatio-temporal modelsdouble generalized linear modelsdispersion modelingcount time seriesGARCHspatial lagquasi-maximum likelihoodR package
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper presents a modeling framework and companion software for autoregressive spatio-temporal data whose marginal distributions come from an exponential-dispersion family. Its central claim is that both the conditional mean and the dispersion (volatility) of the response can be modeled with spatial-lag regression equations of the same generalized-linear-model type, with parameters estimated by alternating quasi-maximum likelihood. If correct, this unifies previously separate classes: Poisson and negative-binomial count models with fixed dispersion, space-time GARCH-type volatility models, and models with over- or underdispersion can be fit, simulated, forecast, and compared in one package. The paper supports the claim with asymptotic heuristics, extensive simulations, and two empirical applications.

Core claim

The core claim is that a spatio-temporal process can be represented by two linked linear predictors using the same spatial-weighting machinery. The mean predictor regresses a link-transformed conditional mean on past values of the predictor and past observations through row-normalized spatial weight matrices; the dispersion predictor does the same for the dispersion parameter, using pseudo-observations obtained from deviance residuals. Because the dispersion model is fit on this error-generating process, the framework yields spatio-temporal GARCH models for Gaussian data and spatio-temporally varying over- or underdispersion for count data. Estimation alternates between the two submodels, tr

What carries the argument

The load-bearing mechanism is the pair of spatial-lag equations (6) and (8) built from row-normalized spatial weight matrices that define zero-, first-, and higher-order neighborhoods. These equations let a small number of autoregressive parameters capture dependence across potentially thousands of locations. The dispersion equation is driven by deviance-based pseudo-observations whose conditional law is approximated as Gamma with fixed dispersion 2, transforming dispersion estimation into a second GLM-type fit and making the two parameter blocks approximately orthogonal.

Load-bearing premise

The analysis stands on the assumption that deviance-based pseudo-observations for count responses behave like Gamma variates with fixed dispersion 2; this is exact for normal and inverse-Gaussian margins but holds for Poisson only when counts are large (the paper suggests an average around 35), so for typical low-count data the estimated dispersion dynamics are systematically underestimated and their standard errors are unreliable.

What would settle it

Simulate a quasi-Poisson spatio-temporal process with a known intercept such that the average count is about 5, a true dispersion autoregressive coefficient of 0.5, and no feedback term. Fit the package's dispersion model and check whether the 90% sandwich interval for the coefficient contains 0.5; the paper's Figure 20 predicts it will not, because the estimate is biased downward at low means. Repeating the exercise at an average count of about 40 should recover the true coefficient.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the framework holds, researchers can estimate spatio-temporal volatility alongside the conditional mean in one procedure rather than detrending and fitting separate mean and variance models.
  • Count-data analysts gain a direct test for spatio-temporal over- or underdispersion; the paper's simulations indicate Wald tests for zero autoregressive dispersion coefficients hold approximate nominal size even at short series lengths.
  • The package's simulation functions allow generating data from a chosen copula and marginal family, making it possible to study how contemporaneous spatial dependence affects inference.
  • For Gaussian marginals, the dispersion equation reproduces a spatio-temporal log-GARCH structure, bringing financial-style volatility modeling into the spatial-lag GLM framework.
  • Users can choose among Poisson, negative binomial, binomial, gamma, inverse-Gaussian, and quasi variants, with AIC, BIC, and QIC provided for model comparison.

Where Pith is reading between the lines

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

  • An untested corollary of the paper's own simulation finding: for low-count data, dispersion estimates from this framework should be interpreted as conservative lower bounds on the strength of volatility dynamics, since autoregressive dispersion coefficients are attenuated when average counts are far below 35.
  • The same Gamma-approximation limitation points to a natural testable extension—replacing deviance pseudo-observations with an exact or REML-type correction for count marginals—which would likely remove the remaining downward bias.
  • Because the package assumes time-invariant spatial weight matrices, extending the machinery to time-varying or data-driven networks would connect it to network time-series problems where spatial structure changes over time.
  • The copula-based data generator could be used to stress-test misspecification: with strong contemporaneous dependence, the quasi-likelihood's conditional-independence assumption may still produce consistent mean parameters, but the coverage of standard errors should degrade; this could be checked by simulation.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper describes an R package, glmSTARMA, for fitting spatio-temporal autoregressive models in a generalized linear framework. The mean process is modeled as a function of past linear predictors, past observations, spatial lags, and covariates. The package extends this to double generalized linear models by allowing the dispersion parameter to follow a similar spatio-temporal autoregressive structure, with estimation alternating between mean and dispersion submodels. The paper includes simulation functionality, estimation with quasi-maximum likelihood, standard errors based on a sandwich formula, model selection criteria, two data examples (rotavirus counts and sea surface temperature anomalies), and a simulation study. The authors position the package as the first to allow simultaneous spatio-temporal modeling of both the conditional mean and the dispersion parameter.

Significance. If the methodology is sound, the package fills a real gap: no existing R package for spatio-temporal data models the dispersion process dynamically in a GLM-type framework, and the examples (space-time GARCH for Gaussian data, over-/underdispersion for counts) are practically relevant. The software appears well engineered, with support for copula-based joint distributions, sparse weight matrices, flexible model order specification, and an open-source implementation with reproducible code. The simulation study is extensive and the authors are transparent about several limitations, which is valuable. However, the central statistical claim—valid estimation and inference for the dispersion process—is weakened by the paper's own finding that the gamma approximation for count pseudo-observations fails at low count levels, and by the explicitly conjectural asymptotic justification.

major comments (4)
  1. [§2.2, Remark 1; §D.2.1, Figure 20] The dispersion model (8) is estimated by treating the deviance-based pseudo-observations d_i,t as Gamma-distributed with fixed dispersion 2 (Remark 1). For Poisson responses this approximation is only reliable when the mean is much larger than the threshold stated in Remark 1. The paper's own simulation (Section D.2.1, Figure 20) shows that for average counts below about 35, the autoregressive dispersion parameters are systematically underestimated and the true value falls outside the empirical 90% interval. Since typical count applications—infectious disease counts, crime counts—often have means well below 35, the package's headline capability of modeling spatio-temporal over-/underdispersion is not reliable in exactly those settings. The text acknowledges this ('only when ... reaches a level around 35') but the abstract and conclusion still present the capability in general terms, and
  2. [§A.2, Eq. (A-5); §5] The asymptotic normality of the QMLE is explicitly conjectured ('we conjecture that the estimators ... are asymptotically normal'), not proved or referenced to a theorem covering this dependent, non-stationary DGLM setting. All reported standard errors, p-values, and the QIC in Section 5 rest on this conjecture. The simulation Q-Q plots provide empirical support for some scenarios, but for generalized Poisson marginals the dispersion estimates are biased (Section D.2), so the inference is not validated in the count case that motivates the package. The paper should either supply a rigorous asymptotic reference, or clearly label the inference as heuristic in the abstract and documentation.
  3. [§4.2 vs §2.2/Remark 4] There is an internal contradiction about which pseudo-observations are used. Section 2.2 defines d_i,t via deviance residuals in Eq. (7), and Remark 4 states the default is 'deviance', while Section 4.2 says the dispersion model is fitted to pseudo-observations 'derived from the Pearson residuals of the mean model.' This matters because the orthogonality of mean and dispersion parameters, cited in Remark 4 and Section A.2 to justify separate iterative estimation, is established for deviance residuals in Smyth (1989). If Pearson residuals are actually used, the orthogonality argument and the estimation theory need to be revisited.
  4. [§6.1, Code output] The statement that Poisson, quasi-Poisson, and negative binomial families 'yield identical point estimates, differing only in standard error calculations' is not generally true when the dispersion parameter of the negative binomial is estimated; the mean estimating equations differ unless the dispersion is fixed. This claim should either be corrected or carefully qualified, otherwise users may be misled about the software's behavior.
minor comments (5)
  1. [§7, Algorithm 1] Typos: 'caputuring' in the conclusion; 'disperson' in Algorithm 1 line 18; check spelling throughout the manuscript.
  2. [§6.2] The directed west/east neighbor matrices have row sums equal to zero for boundary pixels, violating the row-normalization assumption W 1_p = 1_p stated in Section 2.1. The text says the impact is negligible, but a formal justification would be useful because the stability conditions in Appendix A rely on row-normalized matrices.
  3. [§3.3] The 'negbin' sampling method for quasi-Poisson models does not simulate underdispersion and silently falls back to a Poisson distribution. This behavior should be documented explicitly in the function help pages and mentioned in the main text.
  4. [§A.1, Eq. (A-3)] The approximation E(ψ_t) in Eq. (A-3) is used to initialize simulations, but it is only heuristic, especially in the presence of covariates. Consider labeling it as such in the simulation section.
  5. [References] The citation style is inconsistent: 'Maletz et al. (2026)' in Appendix C.1 vs 'Maletz et al. (2024)' elsewhere. Please unify.

Circularity Check

0 steps flagged

No significant circularity; the dispersion-model derivation is externally grounded in Smyth's DGLM theory rather than in the paper's own outputs.

full rationale

The central new capability is the DGLM dispersion model (8), estimated from pseudo-observations d_{i,t} defined in (7). This follows Smyth (1989), an independent external source, not a self-citation or a parameter fitted to the quantity later 'predicted'. The mean model (6) is taken from the authors' prior work (Maletz et al. 2024), but that is an extension/implementation, and the paper's novelty claim is the dispersion layer, which rests on Smyth's orthogonality and deviance-residual theory. The Gamma approximation for d_{i,t} (Remark 1) is a stated assumption with an explicit, self-disclosed limitation: the paper's own simulation study (Section D.2.1, Figure 20) shows that for count means below about 35 the autoregressive dispersion parameters are underestimated. This is a validity/accuracy concern, not a circularity: the approximation is not defined in terms of the fitted dispersion parameters, and the estimation does not reduce to re-inserting the target values. Similarly, the asymptotic normality claim in Section A.2 is explicitly labeled a conjecture, so it is an omitted proof rather than a circular step. Self-citations to Maletz et al. (2024) for PSTARMA, identifiability, and p-value halving are auxiliary and not load-bearing for the DGLM dispersion inference itself. No equation or fitted quantity is equivalent by construction to the paper's headline result; the package's simulations compare estimates against known generator parameters, which is standard validation, not circular reasoning.

Axiom & Free-Parameter Ledger

1 free parameters · 6 axioms · 0 invented entities

The package's inference rests on two conjectured results (stability condition and asymptotic normality) and on the Gamma approximation for pseudo-observations, which the authors' simulations show is only reliable for large count means. These are not proven in the paper and should be treated as assumptions.

free parameters (1)
  • transformation offset c = 1 (default; user-settable via 'const')
    Table 2: h(Y)=log(Y+c·1_p) and softplus/softclipping links use this hyperparameter to stabilize transformations; changing c changes the model and the estimates.
axioms (6)
  • domain assumption Conditional distribution Y_{i,t} | F_{t-1} belongs to the exponential dispersion family (eq. 1)
    All inference and simulation are built on this family; non-exponential dispersion models are outside scope.
  • domain assumption Spatial weight matrices are row-normalized (W^(ℓ)1_p = 1_p) for theory
    Stated after eq. (5); required for stability and estimation. The SST example violates this at boundaries and the paper asserts the effect is negligible.
  • ad hoc to paper Stability condition (A-1)/(A-2) ensures stationarity, ergodicity and moments
    Appendix A.1: presented as a conjecture ('we conjecture that...'); it is used to constrain the parameter space in estimation. No proof is provided.
  • ad hoc to paper QMLE asymptotic normality (A-5) with sandwich covariance holds
    Appendix A.2: explicitly conjectured; this is the basis for summary() standard errors and p-values.
  • domain assumption Gamma(mean=phi, disp=2) approximation for pseudo-observations (deviance/Pearson residuals)
    Remark 1; for Poisson this requires conditional means >3 and in practice around 35 before the autoregressive dispersion estimates are unbiased. Underlies the dispersion model (8).
  • ad hoc to paper Orthogonality of mean and dispersion parameters (Smyth 1989) carries over to dependent spatio-temporal data
    Section A.2 states 'we assume that this property carries over to our context' — enables the alternating estimation of Algorithm 1.

pith-pipeline@v1.3.0-alltime-deepseek · 40273 in / 12790 out tokens · 113126 ms · 2026-08-02T07:52:43.893526+00:00 · methodology

0 comments
read the original abstract

The R package glmSTARMA implements autoregressive models for spatio-temporal data at fixed locations, with time-invariant spatial dependency structure. We rely on generalized linear models methodology and unify several approaches for the analysis of spatial count time series. Such models allow the (conditional) mean of the response to depend on past observations, lagged (conditional) expectations, and covariates. The response can be a continuous or a discrete random variable. Additionally, the package develops inference for double generalized linear models, allowing the dispersion parameter(s) of the marginal distributions to be modeled similarly to the mean process. This is a new capability which introduces, for example, spatio-temporal volatility models, such as space-time GARCH processes, and count time series models with spatio-temporal overdispersion and underdispersion. We provide functions for model estimation, simulation, inference, and prediction. Its use is illustrated by data examples.

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

112 extracted references · 3 canonical work pages

  1. [1]

    Non-Uniform Random Variate Generation , rights =

    Luc Devroye , year =. Non-Uniform Random Variate Generation , rights =. doi:10.1007/978-1-4613-8643-8 , publisher =

    doi:10.1007/978-1-4613-8643-8

  2. [2]

    Water Science and Technology , author =

    Spatio-Temporal Statistical Models for River Monitoring Networks , volume =. Water Science and Technology , author =. 2006 , langid =. doi:10.2166/wst.2006.002 , pages =

    doi:10.2166/wst.2006.002 2006

  3. [3]

    Spatiotemporal Analysis of German Real-Estate Prices , volume =

    Philipp Otto and Wolfgang Schmid , urldate =. Spatiotemporal Analysis of German Real-Estate Prices , volume =. The Annals of Regional Science , shortjournal =. 2018 , langid =. doi:10.1007/s00168-016-0789-y , pages =

    doi:10.1007/s00168-016-0789-y 2018

  4. [4]

    Scotto and Christian H

    Ana Martins and Manuel G. Scotto and Christian H. Weiß and Sónia Gouveia , urldate =. Space-Time Integer-Valued. Electronic Journal of Statistics , shortjournal =. 2023 , keywords =. doi:10.1214/23-EJS2183 , number =

    doi:10.1214/23-ejs2183 2023

  5. [5]

    Gilley and C.F

    R.Kelley Pace and Ronald Barry and Otis W. Gilley and C.F. Sirmans , urldate =. A Method for Spatial–temporal Forecasting with an Application to Real Estate Prices , volume =. International Journal of Forecasting , shortjournal =. 2000 , langid =. doi:10.1016/S0169-2070(99)00047-3 , pages =

    doi:10.1016/s0169-2070(99)00047-3 2000

  6. [6]

    Space–time Modeling of Traffic Flow , volume =

    Yiannis Kamarianakis and Poulicos Prastacos , urldate =. Space–time Modeling of Traffic Flow , volume =. Computers & Geosciences , shortjournal =. 2005 , langid =. doi:10.1016/j.cageo.2004.05.012 , pages =

    doi:10.1016/j.cageo.2004.05.012 2005

  7. [7]

    Spatio-Temporal Count Autoregression , volume =

    Steffen Maletz and Konstantinos Fokianos and Roland Fried , urldate =. Spatio-Temporal Count Autoregression , volume =. Data Science in Science , shortjournal =. 2024 , langid =. doi:10.1080/26941899.2024.2425171 , pages =

    arXiv 2024

  8. [8]

    Paul Elhorst and Silvia Emili , urldate =

    J. Paul Elhorst and Silvia Emili , urldate =. A Spatial Econometric Multivariate Model of Okun's Law , volume =. Regional Science and Urban Economics , shortjournal =. 2022 , langid =. doi:10.1016/j.regsciurbeco.2021.103756 , pages =

    arXiv 2022

  9. [9]

    Short-Term Probabilistic Forecasting of Meso-Scale Near-Surface Urban Temperature Fields , volume =

    Byeongseong Choi and Mario Bergés and Elie Bou-Zeid and Matteo Pozzi , urldate =. Short-Term Probabilistic Forecasting of Meso-Scale Near-Surface Urban Temperature Fields , volume =. Environmental Modelling & Software , shortjournal =. 2021 , langid =. doi:10.1016/j.envsoft.2021.105189 , pages =

    arXiv 2021

  10. [10]

    Clark and Philip M

    Nicholas J. Clark and Philip M. Dixon , urldate =. A Class of Spatially Correlated Self-Exciting Statistical Models , volume =. Spatial Statistics , shortjournal =. 2021 , langid =. doi:10.1016/j.spasta.2021.100493 , pages =

    arXiv 2021

  11. [11]

    A Statistical Framework for the Analysis of Multivariate Infectious Disease Surveillance Counts , volume =

    Leonhard Held and Michael Höhle and Mathias Hofmann , urldate =. A Statistical Framework for the Analysis of Multivariate Infectious Disease Surveillance Counts , volume =. Statistical Modelling , shortjournal =. 2005 , langid =. doi:10.1191/1471082X05st098oa , pages =

    doi:10.1191/1471082x05st098oa 2005

  12. [12]

    A Modelling Approach for Correcting Reporting Delays in Disease Surveillance Data , volume =

    Leonardo S Bastos and Theodoros Economou and Marcelo F C Gomes and Daniel A M Villela and Flavio C Coelho and Oswaldo G Cruz and Oliver Stoner and Trevor Bailey and Claudia T Codeço , urldate =. A Modelling Approach for Correcting Reporting Delays in Disease Surveillance Data , volume =. Statistics in Medicine , shortjournal =. 2019 , langid =. doi:10.100...

    doi:10.1002/sim.8303 2019

  13. [13]

    Assessing the Effect of Containment Measures on the Spatio-Temporal Dynamic of

    Maria Michela Dickson and Giuseppe Espa and Diego Giuliani and Flavio Santi and Lucia Savadori , urldate =. Assessing the Effect of Containment Measures on the Spatio-Temporal Dynamic of. Nonlinear Dynamics , shortjournal =. 2020 , langid =. doi:10.1007/s11071-020-05853-7 , pages =

    doi:10.1007/s11071-020-05853-7 2020

  14. [14]

    A Systematic Review of Spatio-Temporal Statistical Models: Theory, Structure, and Applications , rights =

    Isabella Habereder and Thomas Kneib and Isao Echizen and Timo Spinde , urldate =. A Systematic Review of Spatio-Temporal Statistical Models: Theory, Structure, and Applications , rights =. 2025 , note =. doi:10.48550/ARXIV.2511.00422 , shorttitle =

    Pith/arXiv arXiv doi:10.48550/arxiv.2511.00422 2025

  15. [15]

    Udny Yule , urldate =

    G. Udny Yule , urldate =. Why Do We Sometimes Get Nonsense-Correlations Between Time-Series?--a Study in Sampling and the Nature of Time-Series , volume =. Journal of the Royal Statistical Society , shortjournal =. doi:10.2307/2341482 , shorttitle =

    doi:10.2307/2341482

  16. [16]

    The Summation of Random Causes as the Source of Cyclic Processes , volume =

    Eugen Slutzky , urldate =. The Summation of Random Causes as the Source of Cyclic Processes , volume =. Econometrica , shortjournal =. doi:10.2307/1907241 , pages =

    doi:10.2307/1907241

  17. [17]

    A Study in the Analysis of Stationary Time Series , publisher =

    Herman Wold , year =. A Study in the Analysis of Stationary Time Series , publisher =

  18. [18]

    George E. P. Box and Gwilym M. Jenkins , year =. Time Series Analysis: Forecasting and Control , publisher =

  19. [19]

    A. D. Cliff and J. K. Ord , urldate =. Space-Time Modelling with an Application to Regional Forecasting , doi =. Transactions of the Institute of British Geographers , shortjournal =

  20. [20]

    R. L. Martin and J. E. Oeppen , urldate =. The Identification of Regional Forecasting Models using Space: Time Correlation Functions , doi =. Transactions of the Institute of British Geographers , shortjournal =

  21. [21]

    Pfeifer and Stuart Jay Deutsch , urldate =

    Phillip E. Pfeifer and Stuart Jay Deutsch , urldate =. Identification and Interpretation of First Order Space-Time. Technometrics , shortjournal =. 1980 , langid =. doi:10.1080/00401706.1980.10486172 , pages =

    arXiv 1980

  22. [22]

    Pfeifer , urldate =

    Stuart Jay Deutsch and Phillip E. Pfeifer , urldate =. Space-Time. Technometrics , shortjournal =. 1981 , langid =. doi:10.1080/00401706.1981.10487686 , pages =

    arXiv 1981

  23. [23]

    Pfeifer and Stuart Jay Deutsch , urldate =

    Phillip E. Pfeifer and Stuart Jay Deutsch , urldate =. Variance of the Sample Space-Time Correlation Function of Contemporaneously Correlated Variables , volume =. 1981 , langid =. doi:10.1137/0140010 , pages =

    doi:10.1137/0140010 1981

  24. [24]

    Pfeifer and Stuart Jay Deutsch , urldate =

    Phillip E. Pfeifer and Stuart Jay Deutsch , urldate =. A. Transactions of the Institute of British Geographers , shortjournal =. doi:10.2307/621846 , pages =

    doi:10.2307/621846

  25. [25]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , author =

    Variance of the Sample Space-Time Autocorrelation Function , volume =. Journal of the Royal Statistical Society Series B: Statistical Methodology , author =. 1981 , langid =. doi:10.1111/j.2517-6161.1981.tb01144.x , pages =

    arXiv 1981

  26. [26]

    Pfeifer and Stuart Jay Deutsch , urldate =

    Phillip E. Pfeifer and Stuart Jay Deutsch , urldate =. Seasonal Space‐Time. Geographical Analysis , shortjournal =. 1981 , langid =. doi:10.1111/j.1538-4632.1981.tb00720.x , pages =

    arXiv 1981

  27. [27]

    Sims , urldate =

    Christopher A. Sims , urldate =. Macroeconomics and Reality , volume =. Econometrica , shortjournal =. 1980 , file =. doi:10.2307/1912017 , pages =

    doi:10.2307/1912017 1980

  28. [28]

    G. C. Tiao and G. E. P. Box , urldate =. Modeling Multiple Times Series with Applications , volume =. Journal of the American Statistical Association , shortjournal =. doi:10.2307/2287575 , pages =

    doi:10.2307/2287575

  29. [29]

    Quenouille , year =

    Maurice H. Quenouille , year =. The Analysis of Multiple Time Series , pagetotal =

  30. [30]

    Svetlana Borovkova and Rik Lopuhaa , urldate =. Spatial. 2012 , langid =. doi:10.2139/ssrn.2176781 , shorttitle =

    doi:10.2139/ssrn.2176781 2012

  31. [31]

    A Stationary Spatio‐Temporal

    Sondre Hølleland and Hans Arnfinn Karlsen , urldate =. A Stationary Spatio‐Temporal. Journal of Time Series Analysis , shortjournal =. 2020 , langid =. doi:10.1111/jtsa.12498 , pages =

    doi:10.1111/jtsa.12498 2020

  32. [32]

    Generalised Spatial and Spatiotemporal Autoregressive Conditional Heteroscedasticity , volume =

    Philipp Otto and Wolfgang Schmid and Robert Garthoff , urldate =. Generalised Spatial and Spatiotemporal Autoregressive Conditional Heteroscedasticity , volume =. Spatial Statistics , shortjournal =. 2018 , langid =. doi:10.1016/j.spasta.2018.07.005 , pages =

    doi:10.1016/j.spasta.2018.07.005 2018

  33. [33]

    Lopuhaä and Budi Nurani Ruchjana , urldate =

    Svetlana Borovkova and Hendrik P. Lopuhaä and Budi Nurani Ruchjana , urldate =. Consistency and Asymptotic Normality of Least Squares Estimators in Generalized. Statistica Neerlandica , shortjournal =. 2008 , langid =. doi:10.1111/j.1467-9574.2008.00391.x , pages =

    arXiv 2008

  34. [34]

    Generalized Autoregressive Conditional Heteroskedasticity , volume =

    Tim Bollerslev , urldate =. Generalized Autoregressive Conditional Heteroskedasticity , volume =. Journal of Econometrics , shortjournal =. 1986 , langid =. doi:10.1016/0304-4076(86)90063-1 , pages =

    doi:10.1016/0304-4076(86)90063-1 1986

  35. [35]

    Endemic-Epidemic Models with Discrete-Time Serial Interval Distributions for Infectious Disease Prediction , volume =

    Johannes Bracher and Leonhard Held , urldate =. Endemic-Epidemic Models with Discrete-Time Serial Interval Distributions for Infectious Disease Prediction , volume =. International Journal of Forecasting , shortjournal =. 2022 , langid =. doi:10.1016/j.ijforecast.2020.07.002 , pages =

    doi:10.1016/j.ijforecast.2020.07.002 2022

  36. [36]

    M. A. Al‐Osh and A. A. Alzaid , journal =. 1987 , title =

    1987

  37. [37]

    McKenzie , urldate =

    Ed. McKenzie , urldate =. Some Simple Models for Discrete Variate Time Series , volume =. 1985 , langid =. doi:10.1111/j.1752-1688.1985.tb05379.x , pages =

    arXiv 1985

  38. [38]

    Journal of the Royal Statistical Society Series C: Applied Statistics , author =

    Approximately Linear. Journal of the Royal Statistical Society Series C: Applied Statistics , author =. 2023 , langid =. doi:10.1093/jrsssc/qlad018 , pages =

    doi:10.1093/jrsssc/qlad018 2023

  39. [39]

    Integer‐Valued

    René Ferland and Alain Latour and Driss Oraichi , urldate =. Integer‐Valued. Journal of Time Series Analysis , shortjournal =. 2006 , langid =. doi:10.1111/j.1467-9892.2006.00496.x , pages =

    arXiv 2006

  40. [40]

    Log-Linear Poisson Autoregression , volume =

    Konstantinos Fokianos and Dag Tjøstheim , urldate =. Log-Linear Poisson Autoregression , volume =. Journal of Multivariate Analysis , shortjournal =. 2011 , langid =. doi:10.1016/j.jmva.2010.11.002 , pages =

    doi:10.1016/j.jmva.2010.11.002 2011

  41. [41]

    Multivariate Count Autoregression , volume =

    Konstantinos Fokianos and Bård Støve and Dag Tjøstheim and Paul Doukhan , urldate =. Multivariate Count Autoregression , volume =. Bernoulli , shortjournal =. 2020 , file =. doi:10.3150/19-BEJ1132 , number =

    doi:10.3150/19-bej1132 2020

  42. [42]

    G. E. P. Box and G. C. Tiao , urldate =. Intervention Analysis with Applications to Economic and Environmental Problems , volume =. Journal of the American Statistical Association , shortjournal =. 1975 , langid =. doi:10.1080/01621459.1975.10480264 , pages =

    arXiv 1975

  43. [43]

    Interventions in log-Linear Poisson Autoregression , volume =

    Konstantinos Fokianos and Roland Fried , urldate =. Interventions in log-Linear Poisson Autoregression , volume =. Statistical Modelling , shortjournal =. 2012 , langid =. doi:10.1177/1471082X1201200401 , pages =

    doi:10.1177/1471082x1201200401 2012

  44. [45]

    Spatial Econometrics: Methods and Models , volume =

    Luc Anselin , urldate =. Spatial Econometrics: Methods and Models , volume =. doi:10.1007/978-94-015-7799-1 , series =

    doi:10.1007/978-94-015-7799-1

  45. [46]

    Journal of Time Series Analysis , author =

    A Negative Binomial Integer-Valued. Journal of Time Series Analysis , author =. 2011 , langid =. doi:10.1111/j.1467-9892.2010.00684.x , shorttitle =

    arXiv 2011

  46. [47]

    Inouye and Eunho Yang and Genevera I

    David I. Inouye and Eunho Yang and Genevera I. Allen and Pradeep Ravikumar , urldate =. A Review of Multivariate Distributions for Count Data Derived from the Poisson Distribution , volume =. 2017 , langid =. doi:10.1002/wics.1398 , pages =

    doi:10.1002/wics.1398 2017

  47. [48]

    Spatial Data Science: With Applications in R , doi =

    Edzer Pebesma and Roger Bivand , urldate =. Spatial Data Science: With Applications in R , doi =. 2023 , langid =

    2023

  48. [49]

    doi:10.1201/9781351769723 , author =

    Spatio-Temporal Statistics with. doi:10.1201/9781351769723 , author =

    doi:10.1201/9781351769723

  49. [50]

    Noel A. C. Cressie and Christopher K. Wikle , year =. Statistics for Spatio-Temporal Data , series =

  50. [51]

    W. R. Tobler , urldate =. A Computer Movie Simulating Urban Growth in the Detroit Region , volume =. Economic Geography , shortjournal =. doi:10.2307/143141 , pages =

    doi:10.2307/143141

  51. [52]

    Network Vector Autoregression , volume =

    Xuening Zhu and Rui Pan and Guodong Li and Yuewen Liu and Hansheng Wang , urldate =. Network Vector Autoregression , volume =. The Annals of Statistics , shortjournal =. 2017 , file =. doi:10.1214/16-AOS1476 , number =

    doi:10.1214/16-aos1476 2017

  52. [53]

    An Introduction to Spatial Data Science with

    Luc Anselin , urldate =. An Introduction to Spatial Data Science with. 2024 , langid =. doi:10.1201/9781003274919 , shorttitle =

    doi:10.1201/9781003274919 2024

  53. [54]

    Count Network Autoregression , volume =

    Mirko Armillotta and Konstantinos Fokianos , urldate =. Count Network Autoregression , volume =. Journal of Time Series Analysis , shortjournal =. 2024 , langid =. doi:10.1111/jtsa.12728 , pages =

    doi:10.1111/jtsa.12728 2024

  54. [55]

    Journal of the Royal Statistical Society

    Generalized Linear Models , volume =. Journal of the Royal Statistical Society. Series A (General) , author =. doi:10.2307/2344614 , pages =

    doi:10.2307/2344614

  55. [56]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , author =

    Exponential Dispersion Models , volume =. Journal of the Royal Statistical Society Series B: Statistical Methodology , author =. 1987 , langid =. doi:10.1111/j.2517-6161.1987.tb01685.x , pages =

    arXiv 1987

  56. [57]

    Steffen Maletz and Konstantinos Fokianos and Roland Fried , year =

  57. [58]

    Wagner Barreto-Souza and Luiza S. C. Piancastelli and Konstantinos Fokianos and Hernando Ombao , title =. Journal of Time Series Analysis , volume =. doi:10.1111/jtsa.12838 , year =

    doi:10.1111/jtsa.12838

  58. [59]

    Econometric Theory , author =

    Count and Duration Time Series with Equal Conditional Stochastic and Mean Orders , volume =. Econometric Theory , author =. doi:10.1017/S0266466620000134 , pages =

    doi:10.1017/s0266466620000134

  59. [60]

    Technometrics , author =

    A Three-Stage Iterative Procedure for Space-Time Modeling , volume =. Technometrics , author =. doi:10.2307/1268381 , pages =

    doi:10.2307/1268381

  60. [61]

    The Annals of Statistics , author =

    Strong Consistency of Maximum Quasi-Likelihood Estimators in Generalized Linear Models with Fixed and Adaptive Designs , volume =. The Annals of Statistics , author =

  61. [62]

    Communications in Statistics - Theory and Methods , author =

    Asymptotic Properties of Quasi-Maximum Likelihood Estimates in Generalized Linear Models , volume =. Communications in Statistics - Theory and Methods , author =. 2011 , keywords =. doi:10.1080/03610926.2010.513792 , pages =

    arXiv 2011

  62. [63]

    Statistics , author =

    Asymptotic Properties of Maximum Quasi-Likelihood Estimators in Generalized Linear Models with Adaptive Designs , volume =. Statistics , author =. 2012 , keywords =. doi:10.1080/02331888.2010.543465 , pages =

    arXiv 2012

  63. [64]

    Science in China Series A , author =

    Asymptotic Normality and Strong Consistency of Maximum Quasi-Likelihood Estimates in Generalized Linear Models , volume =. Science in China Series A , author =. 2006 , keywords =. doi:10.1007/s11425-004-5169-x , pages =

    doi:10.1007/s11425-004-5169-x 2006

  64. [65]

    doi:10.1002/9781119313472 , publisher =

    Christian Francq and Jean-Michel Zakoïan , year =. doi:10.1002/9781119313472 , publisher =

    work page doi:10.1002/9781119313472

  65. [66]

    The Annals of Statistics , author =

    Consistency and Asymptotic Normality of the Maximum Likelihood Estimator in Generalized Linear Models , volume =. The Annals of Statistics , author =. 1985 , keywords =. doi:10.1214/aos/1176346597 , pages =

    arXiv 1985

  66. [67]

    Biometrics , volume =

    Wei Pan , title =. Biometrics , volume =. doi:10.1111/j.0006-341X.2001.00120.x , year =

    Pith/arXiv arXiv 2001

  67. [68]

    Akaike , urldate =

    H. Akaike , urldate =. A New Look at the Statistical Model Identification , volume =. 1974 , keywords =. doi:10.1109/TAC.1974.1100705 , pages =

    arXiv 1974

  68. [69]

    The Annals of Statistics , author =

    Estimating the Dimension of a Model , volume =. The Annals of Statistics , author =. 1978 , keywords =. doi:10.1214/aos/1176344136 , pages =

    arXiv 1978

  69. [70]

    Algorithm 733:

    Dieter Kraft , urldate =. Algorithm 733:. 1994 , keywords =. doi:10.1145/192115.192124 , shorttitle =

    arXiv 1994

  70. [71]

    Johnson , year =

    Jelma Ypma and Steven G. Johnson , year =

  71. [72]

    Journal of Economic Surveys , author =

    Spatial and Spatiotemporal Volatility Models: A Review , volume =. Journal of Economic Surveys , author =. 2025 , keywords =. doi:10.1111/joes.12643 , shorttitle =

    doi:10.1111/joes.12643 2025

  72. [73]

    Journal of the American Statistical Association , author =

    Estimation and Identification of Space-Time. Journal of the American Statistical Association , author =. 1986 , file =. doi:10.2307/2289008 , pages =

    doi:10.2307/2289008 1986

  73. [74]

    doi:10.32614/RJ-2019-053 , pages =

    The R Journal , author =. doi:10.32614/RJ-2019-053 , pages =

    work page doi:10.32614/rj-2019-053 2019

  74. [75]

    Wood , location =

    Simon N. Wood , location =. Generalized Additive Models: An Introduction with. doi:10.1201/9781315370279 , shorttitle =

    doi:10.1201/9781315370279

  75. [76]

    Felix Cheysson , year =

  76. [77]

    Journal of Statistical Software , author =

    Spatio-Temporal Analysis of Epidemic Phenomena Using the. Journal of Statistical Software , author =. 2017 , file =. doi:10.18637/jss.v077.i11 , pages =

    doi:10.18637/jss.v077.i11 2017

  77. [78]

    Statistics in Medicine , author =

    Multivariate Modelling of Infectious Disease Surveillance Data , volume =. Statistics in Medicine , author =. doi:10.1002/sim.3440 , pages =

    doi:10.1002/sim.3440

  78. [79]

    Douglas Bates and Martin Maechler and Mikael Jagan , year =

  79. [80]

    Roger Bivand and David W. S. Wong , year =. Comparing Implementations of Global and Local Indicators of Spatial Association , volume =. doi:10.1007/s11749-018-0599-x , pages =

    doi:10.1007/s11749-018-0599-x

  80. [81]

    Spatial Statistics for Data Science: Theory and Practice with

    Paula Moraga , year =. Spatial Statistics for Data Science: Theory and Practice with

Showing first 80 references.