pith. sign in

arxiv: 2604.13779 · v2 · pith:MDURBTQQnew · submitted 2026-04-15 · 🧮 math.ST · stat.TH

The Integer-valued Moving-Average Random Field

Angelika Silbernagel , Christian H. Wei{\ss} This is my paper

Pith reviewed 2026-05-25 07:01 UTC · model grok-4.3

classification 🧮 math.ST stat.TH
keywords integer-valued moving averagerandom fieldcount dataspatial dependencePoisson marginalautocovariancebivariate distributionspatial statistics
0
0 comments X

The pith

An integer-valued moving-average random field yields closed-form marginal distributions and spatial autocovariances for arbitrary order including multilateral cases.

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

The paper proposes an INMA model for count-valued random fields on a spatial lattice. It derives closed-form expressions for the marginal distribution of each site and for the full spatial dependence structure, covering any finite order and both unilateral and multilateral neighborhoods. General formulas are given for the bivariate joint distributions and for the autocovariances, and the construction is shown to support Poisson margins among other discrete distributions. A sympathetic reader cares because the closed forms make exact likelihood evaluation and moment calculations feasible for spatial count data without simulation or approximation. The paper also illustrates that the model can reproduce a variety of interpretable dependence patterns and can approximate a target spatial covariance structure.

Core claim

The INMA random field is constructed as a spatial moving average of an underlying integer-valued innovation process; closed-form expressions are derived for its marginal distribution and spatial dependence structure for arbitrary model order and also covering the multilateral case, with general expressions supplied for bivariate distributions and autocovariances, and the field can be equipped with a Poisson marginal distribution while permitting different well-interpretable dependence structures.

What carries the argument

The integer-valued moving-average (INMA) random field, a linear filter of an innovation field over a spatial neighborhood whose coefficients and innovation distribution together produce closed-form marginals and joints.

If this is right

  • The model permits exact computation of probabilities and moments for spatial count observations at any lattice site.
  • Different spatial dependence patterns can be realized and interpreted directly from the choice of coefficients and neighborhood.
  • The construction supplies an INMA approximation to any prescribed spatial dependence structure for count data.
  • Real-data fitting becomes feasible via the closed-form bivariate distributions without requiring numerical integration.

Where Pith is reading between the lines

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

  • Extensions to other discrete marginals such as negative binomial could be obtained by the same filtering construction.
  • Direct comparison with existing spatial count models would clarify when the closed-form advantage outweighs the linear-filter restriction.
  • Applications to lattice data in ecology or epidemiology could use the exact autocovariance expressions for model selection.
  • The multilateral case opens the possibility of isotropic dependence on regular grids that is difficult to achieve with many other integer spatial models.

Load-bearing premise

The innovation process and the moving-average coefficients must allow closed-form expressions for the marginal and joint distributions at arbitrary orders and multilateral neighborhoods without further restrictions that would invalidate the formulas.

What would settle it

Explicit construction of a specific bilateral order-2 INMA field whose bivariate distribution cannot be expressed in closed form would falsify the general claim.

Figures

Figures reproduced from arXiv: 2604.13779 by Angelika Silbernagel, Christian H. Wei{\ss}.

Figure 1
Figure 1. Figure 1: Illustration of the regions for the innovations appearing in Xs,t (dashed) and Xs−k,t−l (dotted). The region containing the overlap is highlighted in gray. · q Y1+k i=q1+1 q Y2+l j=l pgfε  1 + βi−k,j−l (u2 − 1) · Y q1 i=k q Y2+l j=q2+1 pgfε  1 + βi−k,j−l (u2 − 1) · Y q1 i=k Y q2 j=l pgfε  1 + βij (u1 − 1) + βi−k,j−l (u2 − 1) + P(Z (i,j) s,t;1 = Z (i−k,j−l) s,t;1 = 1) · (u1 − 1)(u2 − 1) , which covers… view at source ↗
read the original abstract

An integer-valued moving average (INMA) model for count random fields is proposed and investigated. Closed-form expressions are derived for both its marginal distribution and spatial dependence structure, for arbitrary model order and also covering the multilateral case. In particular, general expressions for bivariate distributions and autocovariances are provided. It is shown that the INMA random field can be equipped (among others) with a Poisson marginal distribution. It is also demonstrated that different and well-interpretable dependence structures are possible. For illustration, we discuss a real-world data example and propose an INMA approximation to a given spatial dependence structure.

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

1 major / 0 minor

Summary. The paper proposes an integer-valued moving-average (INMA) model for count random fields. It derives closed-form expressions for the marginal distribution and spatial dependence structure (including bivariate distributions and autocovariances) that hold for arbitrary model order and cover the multilateral case. The model is shown to admit Poisson marginals (among others), to support different interpretable dependence structures, and is illustrated via a real-data example plus an approximation method for a target spatial dependence structure.

Significance. If the claimed closed-form expressions are valid without hidden restrictions on innovations or coefficients, the work would supply a tractable parametric family for spatial count data with explicit marginals and dependence, extending univariate INMA models to random fields. The Poisson-marginal case and the multilateral coverage would be particularly useful for applications in spatial statistics.

major comments (1)
  1. [Abstract] Abstract and §1: the central claim that closed-form expressions exist for marginals, bivariate distributions, and autocovariances at arbitrary order (including multilateral) is load-bearing, yet the manuscript provides no explicit statement of the required conditions on the innovation distribution or on the moving-average coefficients that keep the convolutions or overlaps in closed form. Without these restrictions the general formulas cannot hold for every order and dependence structure.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed reading and the constructive comment on the central claims of the paper. We address the point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract and §1: the central claim that closed-form expressions exist for marginals, bivariate distributions, and autocovariances at arbitrary order (including multilateral) is load-bearing, yet the manuscript provides no explicit statement of the required conditions on the innovation distribution or on the moving-average coefficients that keep the convolutions or overlaps in closed form. Without these restrictions the general formulas cannot hold for every order and dependence structure.

    Authors: We agree that an explicit statement of the modeling assumptions is needed to make the scope of the closed-form results fully transparent. The INMA random field is defined with i.i.d. non-negative integer-valued innovations and non-negative integer moving-average coefficients; the marginal and bivariate distributions are then obtained via finite convolutions whose closed-form character holds for innovation families closed under convolution (Poisson, negative binomial, etc.) and, more generally, can be expressed in terms of the innovation probability mass function. The multilateral case is covered by the same convolution structure once the support of the coefficient array is fixed. In the revised manuscript we will insert a dedicated paragraph (new §2.2) that states these conditions, together with a short remark in the abstract and the opening of §1, and we will add a sentence clarifying that the formulas remain valid for arbitrary order provided the coefficient array is finite. revision: yes

Circularity Check

0 steps flagged

No circularity: derivations are direct from model definition without reduction to fits or self-citations.

full rationale

The paper proposes an INMA model and derives closed-form expressions for marginals, bivariate distributions, and autocovariances directly from the linear combination of innovations and coefficients. No evidence in the provided abstract or reader's summary of self-definitional loops, fitted parameters renamed as predictions, or load-bearing self-citations. The Poisson marginal claim follows from standard properties of independent Poisson innovations under non-negative coefficients, which is a standard construction rather than a circular renaming. The central claims remain independent of any fitted inputs or prior author results invoked as uniqueness theorems. This is the expected outcome for a model-definition paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no explicit free parameters, axioms, or invented entities are identifiable; the model appears to rest on standard assumptions for moving-average processes that are not detailed here.

pith-pipeline@v0.9.0 · 5621 in / 1046 out tokens · 40834 ms · 2026-05-25T07:01:22.567184+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

19 extracted references · 19 canonical work pages

  1. [1]

    Sabyasachi Basu and Gregory C Reinsel

    doi: 10.1007/BF02924535. Sabyasachi Basu and Gregory C Reinsel. Properties of the spatial unilateral first-order ARMA model.Advances in Applied Probability, 25(3):631–648,

    work page doi:10.1007/bf02924535

  2. [2]

    Julian Besag

    doi: 10.2307/1427527. Julian Besag. Spatial interaction and the statistical analysis of lattice systems.Journal of the Royal Statistical Society: Series B (Methodological), 36(2):192–225,

    work page doi:10.2307/1427527

  3. [3]

    1974.tb00999.x

    doi: 10.1111/j.2517-6161. 1974.tb00999.x. Kurt Br¨ ann¨ as and Andreia Hall. Estimation in integer-valued moving average models.Applied Stochastic Models in Business and Industry, 17(3):277–291,

    work page doi:10.1111/j.2517-6161 1974

  4. [4]

    Azmi Chutoo, Dimitris Karlis, Naushad Mamode Khan, and Vandna Jowaheer

    doi: 10.1002/asmb.445. Azmi Chutoo, Dimitris Karlis, Naushad Mamode Khan, and Vandna Jowaheer. The unilateral spatial autogressive process for the regular lattice two-dimensional spatial discrete data.SORT- Statistics and Operations Research Transactions, 45(1):67–90,

    work page doi:10.1002/asmb.445

  5. [5]

    Alireza Ghodsi

    doi: 10.2436/20.8080.02.110. Alireza Ghodsi. Conditional maximum likelihood estimation of the first-order spatial integer-valued autoregressive (SINAR(1, 1)) model.Journal of the Iranian Statistical Society, 14(2):15–36,

    work page doi:10.2436/20.8080.02.110

  6. [6]

    Alireza Ghodsi and Hassan S

    doi: 10.7508/jirss.2015.02.002. Alireza Ghodsi and Hassan S. Bakouch. Spatial INAR(1,1) model based on mixing Pegram and binomial thinning operators with fitting striga counts.Communications in Statistics - Theory and Methods, 54(21):6988–6996,

    work page doi:10.7508/jirss.2015.02.002 2015

  7. [7]

    Alireza Ghodsi, Mahendran Shitan, and Hassan S Bakouch

    doi: 10.1080/03610926.2025.2465647. Alireza Ghodsi, Mahendran Shitan, and Hassan S Bakouch. A first-order spatial integer-valued autoregressive SINAR(1,

    work page doi:10.1080/03610926.2025.2465647 2025

  8. [8]

    Alireza Ghodsi, Hassan S

    doi: 10.1080/03610926.2011.560739. Alireza Ghodsi, Hassan S. Bakouch, and Mahendran Shitan. First-order spatial dependent count integer-valued autoregressive (Sp-DCINAR(1,1)) process.Communications in Statistics - Simu- lation and Computation, 53(12):6050–6060,

    work page doi:10.1080/03610926.2011.560739 2011

  9. [9]

    Robert P

    doi: 10.1080/03610918.2023.2234681. Robert P. Haining. The moving average model for spatial interaction.Transactions of the Institute of British Geographers, 3(2):202–225,

    work page doi:10.1080/03610918.2023.2234681 2023

  10. [10]

    Dimitris Karlis, Azmi Chutoo, Naushad Mamode Khan, and Vandna Jowaheer

    doi: 10.2307/622202. Dimitris Karlis, Azmi Chutoo, Naushad Mamode Khan, and Vandna Jowaheer. The multilateral spatial integer-valued process of order 1.Statistica Neerlandica, 78(1):4–24,

    work page doi:10.2307/622202

  11. [11]

    Gilberto Pereira Sassi and Carolina Costa Mota Para´ ıba

    doi: 10.2307/1427362. Gilberto Pereira Sassi and Carolina Costa Mota Para´ ıba. Conditional least squares estimation for the SINAR(1,

    work page doi:10.2307/1427362

  12. [12]

    Angelika Silbernagel and Christian H

    doi: 10.1080/03610918.2020.1871489. Angelika Silbernagel and Christian H. Weiß. The autocorrelation structure of integer-valued au- toregressive random fields.Statistics & Probability Letters, 233:110681,

    work page doi:10.1080/03610918.2020.1871489 2020

  13. [13]

    2026.110681

    doi: 10.1016/j.spl. 2026.110681. 11 Atefeh Tabandeh and Alireza Ghodsi. First-order spatial random coefficient non-negative integer- valued autoregressive (SRCINAR(1,1)) model.Communications in Statistics - Simulation and Computation, 53(6):2662–2674,

    work page doi:10.1016/j.spl 2026

  14. [14]

    Christian H Weiß

    doi: 10.1080/03610918.2022.2083164. Christian H Weiß. Serial dependence and regression of Poisson INARMA models.Journal of Statistical Planning and Inference, 138(10):2975–2990,

    work page doi:10.1080/03610918.2022.2083164 2022

  15. [15]

    Christian H Weiß.An Introduction to Discrete-Valued Time Series

    doi: 10.1016/j.jspi.2007.11.009. Christian H Weiß.An Introduction to Discrete-Valued Time Series. John Wiley & Sons, Ltd, 1st ed edition,

    work page doi:10.1016/j.jspi.2007.11.009 2007

  16. [16]

    doi: 10.1016/j.spasta.2023.100800. P. Whittle. On stationary processes in the plane.Biometrika, 41(3–4):434–449,

    work page doi:10.1016/j.spasta.2023.100800 2023

  17. [17]

    doi: 10.1080/ 02664763.2025.2565593. AppendixA.Derivations A.1.Proof of Proposition 2.1.The first expression is an immediate consequence of the fact that for all constantsα∈[0,1] and random variablesX, it holdsα◦X|X∼Bin(X,α) and hence, by the law of total expectation,E(α◦X) =E(E(α◦X|X)) =αEX. Since (εs,t) is i.i.d., and since the thinnings applied to diff...

    work page arXiv 2025

  18. [18]

    Note that for a random variableXand a constantα∈[0,1], it holds V(α◦X) =V(E(α◦X|X)) +E(V(α◦X|X)) =V(α·X) +α(1−α)EX̸=V(α·X), see Weiß [2018, p

    implies V(Xs,t) = q1∑ i=0 q2∑ j=0 V(βij◦εs−i,t−j) = q1∑ i=0 q2∑ j=0 ( β2 ijσ2 ε+βij(1−βij)µε ) , which leads to the desired expression. Note that for a random variableXand a constantα∈[0,1], it holds V(α◦X) =V(E(α◦X|X)) +E(V(α◦X|X)) =V(α·X) +α(1−α)EX̸=V(α·X), see Weiß [2018, p. 17], for instance. With regard to the pgf, first we consider pgf β◦ε(u) =E(u β...

    work page 2018

  19. [19]

    ) = exp ( µX(u1 +u 2−2) ) ·exp ( µX(u1−1)(u2−1)ρ(k,l) ) . The proof of the conditional moments in Corollary 4.2 (iii) goes along the same lines as that of Theorem 3.1 in Weiß [2008], whereρ(k) is replaced byρ(k,l), and where the definition ofβ •is adjusted accordingly, because aside from the aforementioned terms, the pgfs under consideration 14 A. SILBERN...

    work page 2008