pith. sign in

arxiv: 2605.00470 · v1 · submitted 2026-05-01 · 📊 stat.ME · stat.AP

Robust spatial scalar-on-function regression: A Fisher-consistent redescending M-estimation approach

Muge Mutis , Ufuk Beyaztas , Han Lin Shang This is my paper

Pith reviewed 2026-05-09 19:26 UTC · model grok-4.3

classification 📊 stat.ME stat.AP
keywords robust estimationspatial scalar-on-function regressionM-estimationFisher consistencyredescending loss functionsfunctional principal component analysisspatial autoregressive model
0
0 comments X

The pith

A Fisher-consistent redescending M-estimator provides robust estimates for spatial scalar-on-function regression models under contamination.

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

The paper develops a new estimation method for models where a scalar outcome is predicted by a functional variable and a spatial lag term. Standard approaches can fail with outliers or strong spatial effects, but this method first uses robust principal components to reduce the functional data dimension cleanly, then solves a system of equations that remain consistent even with redescending loss functions. It jointly finds the coefficients, the spatial parameter, and a scale factor. Simulations and real air-quality data show it outperforms existing robust and classical methods when data are contaminated. The theoretical results include consistency and asymptotic normality for the estimates and the recovered slope function.

Core claim

We develop a Fisher-consistent redescending robust estimator for the spatial scalar-on-function regression model, where a scalar response depends on both a functional predictor and a spatial autoregressive lag. The estimator applies robust functional principal component analysis to obtain a contamination-resistant finite-dimensional representation of the functional predictor and then estimates the resulting spatial regression model through a bias-corrected system of M-estimating equations that allow redescending loss functions. The method jointly estimates the regression coefficients, spatial dependence parameter, and scale parameter within a unified framework, supported by a hybrid IRLS-New

What carries the argument

Bias-corrected system of M-estimating equations in a Fisher-consistent framework that jointly estimates regression coefficients, spatial dependence parameter, and scale parameter, after robust functional principal component analysis reduces the functional predictor to a finite-dimensional representation.

Load-bearing premise

The functional predictor admits a low-dimensional representation via robust functional principal component analysis, and the spatial dependence is adequately captured by a scalar autoregressive lag term under standard regularity conditions for M-estimation.

What would settle it

Monte Carlo experiments or a real dataset where the proposed estimators show no improvement or worse performance than classical or Huber-type methods under heavy contamination with vertical outliers and leverage points would falsify the superiority under contamination.

Figures

Figures reproduced from arXiv: 2605.00470 by Han Lin Shang, Muge Mutis, Ufuk Beyaztas.

Figure 1
Figure 1. Figure 1: Flowchart of the proposed FCSAR estimation algorithm. view at source ↗
Figure 2
Figure 2. Figure 2: Plots of the generated true regression coefficient functions and mean functions of the estimated view at source ↗
Figure 3
Figure 3. Figure 3: Plots of the generated true regression coefficient functions (black lines), estimated coefficient view at source ↗
Figure 4
Figure 4. Figure 4: Spatial distribution of the 202 air-quality monitoring stations used in the empirical analysis across view at source ↗
Figure 5
Figure 5. Figure 5: Observed air-pollution measurements from the 202 monitoring stations in France. The top panels view at source ↗
Figure 6
Figure 6. Figure 6: Local Moran’s I scatter plots for the 2022 annual average PM10 concentrations. The horizontal axis shows the centered PM10 values, while the vertical axis shows their spatial lag based on three alternative row-standardized weight matrices: the inverse-distance weighting WIDW (left panel), the k-nearest neighbors WkNN (middle panel), and the distance-band WDB (right panel). We fit the six competing procedur… view at source ↗
Figure 7
Figure 7. Figure 7: Estimated regression coefficient functions view at source ↗
read the original abstract

We develop a Fisher-consistent redescending robust estimator for the spatial scalar-on-function regression model, where a scalar response depends on both a functional predictor and a spatial autoregressive lag. Existing estimation procedures for this model are typically based on likelihood methods or monotone-loss robust M-estimators. They may be highly sensitive to vertical outliers, leverage points in the functional predictor, and numerical instability induced by strong spatial dependence. To address these issues, we propose a new estimation framework that first applies robust functional principal component analysis to obtain a contamination-resistant finite-dimensional representation of the functional predictor and then estimates the resulting spatial regression model through a bias-corrected system of M-estimating equations. The proposed method allows redescending loss functions, including Andrews' sine and Danish losses, and jointly estimates the regression coefficients, spatial dependence parameter, and scale parameter within a unified Fisher-consistent framework. For computation, we develop a hybrid IRLS-Newton algorithm that combines weighted least-squares updates for the regression parameters with a Newton-Raphson update for the spatial parameter. We establish Fisher consistency, consistency, asymptotic normality, and the asymptotic distribution of the reconstructed slope function. Monte Carlo experiments show that the proposed estimators remain competitive under clean data and substantially outperform classical and Huber-type robust competitors under contamination, particularly in severe outlier settings. An application to French air-quality data further demonstrates improved predictive performance and stable estimation of spatial dependence. Our method has been implemented in the fcsar R package.

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 / 1 minor

Summary. The manuscript develops a Fisher-consistent redescending robust estimator for spatial scalar-on-function regression using robust FPCA for the functional predictor followed by a bias-corrected M-estimation framework that accommodates redescending loss functions. It establishes Fisher consistency, consistency, asymptotic normality, and the asymptotic distribution of the reconstructed slope function, with supporting Monte Carlo experiments and a real-data application.

Significance. If the claimed Fisher consistency holds in this setting with spatial dependence and redescending losses, the approach would provide a useful robust alternative to standard methods, particularly for contaminated functional data with spatial structure. The R package implementation and empirical results add to its practical value.

major comments (1)
  1. [§3] The bias correction in the system of M-estimating equations must account for the observation-specific weights from the redescending loss function when there is a spatial autoregressive lag term. The provided description does not clarify whether the correction term incorporates these weights or assumes constant weights; without this, Fisher consistency may not hold as the expectation of the estimating function at the true parameter could be nonzero. Please provide the explicit derivation and verification of the zero-expectation property.
minor comments (1)
  1. [Abstract] The hybrid IRLS-Newton algorithm is mentioned but lacks a detailed algorithmic description or pseudocode, which would aid implementation and understanding.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thorough review and valuable feedback on our manuscript. We address the major comment in detail below, providing the requested clarification on the bias correction procedure.

read point-by-point responses

  1. Referee: [§3] The bias correction in the system of M-estimating equations must account for the observation-specific weights from the redescending loss function when there is a spatial autoregressive lag term. The provided description does not clarify whether the correction term incorporates these weights or assumes constant weights; without this, Fisher consistency may not hold as the expectation of the estimating function at the true parameter could be nonzero. Please provide the explicit derivation and verification of the zero-expectation property.

    Authors: We are grateful for this observation, as it allows us to clarify an important technical detail. In the development of the bias-corrected M-estimating equations in Section 3, the correction term is explicitly constructed to incorporate the observation-specific weights from the redescending loss function. The estimating equations are formulated such that the bias correction is the expectation of the weighted psi-function applied to the residuals, conditional on the functional principal components and accounting for the spatial autoregressive term. This ensures that the expectation of the estimating function evaluated at the true parameter values is zero, even in the presence of spatial dependence and variable weights, thereby establishing Fisher consistency. We acknowledge that the original manuscript description could have been more explicit on this point. In the revised version, we will provide the full derivation of the zero-expectation property, including the steps verifying that the weighted correction term nullifies the expectation under the model assumptions with the spatial lag. This will be presented with mathematical detail to confirm the property holds. revision: yes

Circularity Check

0 steps flagged

Minor self-citation in robust FPCA and M-estimation components; central Fisher-consistency claim remains independent

full rationale

The derivation applies standard robust FPCA for dimension reduction of the functional predictor followed by a bias-corrected system of M-estimating equations using redescending losses. Fisher consistency, consistency, and asymptotic normality are asserted to hold under regularity conditions for M-estimation with the spatial lag term. No equation reduces the target slope function or spatial parameter to a quantity defined by the same estimator, and no self-citation chain is load-bearing for the consistency proof. The framework builds on established robust statistics tools without renaming or smuggling ansatzes via self-reference.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that the spatial scalar-on-function model is correctly specified and that robust FPCA yields a faithful finite-dimensional representation; no new entities are postulated.

axioms (2)
  • domain assumption The observed data follow a spatial scalar-on-function regression model with a scalar response, functional predictor, and autoregressive spatial lag.
    This is the model class for which the estimator is derived.
  • standard math Standard regularity conditions for M-estimation and asymptotic normality hold.
    Invoked to obtain the stated consistency and normality results.

pith-pipeline@v0.9.0 · 5569 in / 1377 out tokens · 34834 ms · 2026-05-09T19:26:31.656453+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [1]

    (1998),Spatial Econometrics: Methods and Models, Springer, Berlin

    Anselin, L. (1998),Spatial Econometrics: Methods and Models, Springer, Berlin

    work page 1998

  2. [2]

    Bali, J. L. and Boente, G. (2009), ‘Principal points and elliptical distributions from the multivariate setting to the functional case’,Statistics & Probability Letters79(17), 1858–1865

    work page 2009

  3. [3]

    Bali, J. L. and Boente, G. (2015), ‘Influence function of projection-pursuit principal components for functional data’,Journal of Multivariate Analysis133, 173–199

    work page 2015

  4. [4]

    L., Boente, G., Tyler, D

    Bali, J. L., Boente, G., Tyler, D. E. and Wang, J.-L. (2011), ‘Robust functional principal components: A projection-pursuit approach’,The Annals of Statistics39(6), 2852–2882

    work page 2011

  5. [5]

    and Shang, H

    Beyaztas, U., Mandal, A. and Shang, H. L. (2026), ‘Enhancing spatial functional linear regression with robust dimension reduction methods’,Journal of Multivariate Analysis211, 105538

    work page 2026

  6. [6]

    and Nkiet, G

    Bouka, S., Dabo-Niang, S. and Nkiet, G. M. (2023), ‘On estimation and prediction in spatial functional linear regression model’,Lithuanian Mathematical Journal63(1), 13–30

    work page 2023

  7. [7]

    Cliff, A. D. and Ord, J. K. (1973),Spatial Autocorrelation, Pion, London

    work page 1973

  8. [8]

    and Genton, M

    Dai, W. and Genton, M. G. (2018), ‘Multivariate functional data visualization and outlier detection’, Journal of Computational and Graphical Statistics27(4), 923–934

    work page 2018

  9. [9]

    and Koch, W

    Ertur, C. and Koch, W. (2007), ‘Growth, technological interdependence and spatial externalities: Theory and evidence’,Journal of Applied Econometrics22(6), 1033–1062

    work page 2007

  10. [10]

    Genton, M. G. and Ronchetti, E. (2003), ‘Robust indirect inference’,Journal of the American Statistical Association98(461), 67–76

    work page 2003

  11. [11]

    Grange, S. K. (2019),Technical note: saqgetr R package. URL: https://drive.google.com/open?id=1IgDODHqBHewCTKLdAAxRyR7ml8ht6Ods

    work page 2019

  12. [12]

    and Mallows, C

    Hastie, T. and Mallows, C. (1993), ‘A statistical view of some chemometrics regression tools: Discussion’,Technometrics35(2), 140–143

    work page 1993

  13. [13]

    J., Li, Y

    Hu, X. J., Li, Y. L. and Shi, X. P . (2021), ‘Functional spatial autoregressive analysis of SO2 and air temperature in Fenwei Plain’,Science Technology and Engineering21(28), 11938–11946

    work page 2021

  14. [14]

    and Wang, S

    Huang, T., Saporta, G., Wang, H. and Wang, S. (2021), ‘A robust spatial autoregressive scalar-on- function regression with t-distribution’,Advances in Data Analysis and Classification15(1), 57–81

    work page 2021

  15. [15]

    Luna, X. D. and Genton, M. G. (2002), ‘Simulation-based inference for simultaneous processes on regular lattices’,Statistics and Computing12(2), 125–134

    work page 2002

  16. [16]

    and Wang, H

    Ma, Y., Pan, R., Zou, T. and Wang, H. (2020), ‘A naive least squares method for spatial autoregres- sion with covariates’,Statistica Sinica30(2), 653–672. 50

    work page 2020

  17. [17]

    T., Lillo, R

    Ojo, O. T., Lillo, R. E. and Anta, A. F. (2023),fdaoutlier: Outlier Detection Tools for Functional Data Analysis. R package version 0.2.1. URL:https://CRAN.R-project.org/package=fdaoutlier

    work page 2023

  18. [18]

    and Porcu, E

    Pineda-Rios, W., Giraldo, R. and Porcu, E. (2019), ‘Functional SAR models: with application to spatial econometrics’,Spatial Statistics29, 145–159

    work page 2019

  19. [19]

    Ruiz-Medina, M. D. (2011), ‘Spatial autoregressive and moving average Hilbertian processes’, Journal of Multivariate Analysis102(2), 292–305

    work page 2011

  20. [20]

    Ruiz-Medina, M. D. (2012), ‘Spatial functional prediction from spatial autoregressive Hilbertian processes’,Environmetrics23(1), 119–128

    work page 2012

  21. [21]

    Y., Ding, D., Hui, F

    Tho, Z. Y., Ding, D., Hui, F. K. C., Welsh, A. H. and Zou, T. (2023), ‘On the robust estimation of spatial autoregressive models’,Econometrics and Statisticsin press

    work page 2023

  22. [22]

    and Fang, X

    Zhan, W., Hu, Y., Zeng, W. and Fang, X. (2026), ‘A robust M-estimation framework for spatial autoregressive models: loss function design and optimization strategies’,International Journal of Geographical Information Science40(4), 1246–1270. 51

    work page 2026