betaFeedback

WorksMathematics

Penalized Likelihood Inference in Highly Multivariate Sparse Spectral Graphical Gaussian Processes

Abstract

We develop a penalized likelihood inference approach for highly multivariate Gaussian stochastic processes. Since zeros in the inverse cross-spectral density encode process-level conditional independence, sparse modeling of this object provides a natural route to scalable inference in the highly multivariate case. We introduce a latent variable representation which, combined with random Fourier features, yields a stochastic approximation expectation–maximization algorithm for penalized maximum likelihood estimation of a sparse inverse cross-spectral density matrix under a graphical LASSO penalty. The augmentation is crucial because sparsity is present in the inverse spectral density of the latent variable, whereas the marginal inverse covariance remains dense in the presence of a spatial nugget. Our frequency domain formulation enables joint penalized likelihood estimation of process-specific spatial covariance parameters and sparse process-level conditional dependence without requiring a pre-specified reference set of locations. The method is demonstrated on both synthetic and the MERRA-2 aerosol reanalysis data sets, yielding benefit over existing alternatives, especially for prediction at unobserved locations under spatial misalignment.

Introduction

Construction of highly multivariate cross-covariances is of great contemporary interest in disparate applications, including environmental sciences, where recent technological advances have enabled measurements of multiple correlated variables over a spatial domain, and joint modeling is desired. One example is NASA’s MERRA-2 reanalysis, which incorporates observations from satellite instruments and provides global outputs of atmospheric variables at multiple vertical levels ([21]). Statistical models have been used to study dependence in MERRA-2 fields ([42, 56, 45]), and jointly predicting multiple MERRA-2 fields is also of interest ([26]).

However, flexible approaches to multivariate cross-covariances ([24, 3]) unfortunately do not efficiently scale to a number of variables beyond single digits while retaining flexibility for the individual processes. As evident from the comprehensive review by Genton and Kleiber [22], a key reason is that the traditional approaches construct a fully dense cross-covariance, and there is no notion of sparsity, in a suitable sense to be defined, that enables statistical estimation with respect to three main challenges: managing computation, reducing memory usage, and handling model parsimony and interpretability. The estimation of a fully dense highly multivariate cross-covariance is infeasible both computationally and statistically. In particular, likelihood estimation requires the repeated inversion of a matrix of size np×npnp \times np, where pp is the number of variables or processes, and nn is the average number of data locations per process, incurring a nominal computational cost of O(n3p3)O(n^{3}p^{3}) floating point operations per likelihood evaluation, without further simplifications. Furthermore, memory storage of O(n2p2)O(n^{2}p^{2}) is needed. Finally, the number of parameters grows quadratically in pp, complicating numerical optimization of the likelihood for large pp, and proposing a model that is both flexible and valid can be challenging. The large nn problem has traditionally received attention in the Gaussian process (GP) literature ([50, 15, 28]), as have the problems with asymmetry (see, e.g., [52] (Li and Zhang 2011)) or nonstationarity (see, e.g., [29]) in cross-covariances; whereas the focus on the large pp problem is still at a relatively nascent stage, which we review below.

A simple and traditional large pp approach is the linear model of coregionalization ([25]), which leads to quite restrictive smoothness properties of the component processes, in that the smoothness of any component defaults to that of the roughest ([22]). More flexibly, Dey et al. [17] introduced a notion of stitching processes together at specific spatial locations, where conditional independence is enforced between different processes at other locations using an iterative proportional scaling approach ([47]). Similarly, Peruzzi [40] uses the joint dependence at a reference set of locations to extend the process to the entire spatial domain. A potential disadvantage of reference set-based approaches is that estimates, as well as out-of-sample prediction, may depend on the reference set used for construction, including its ordering, i.e., the reference set becomes part of the model construction. If all model parameters are held fixed but two different reference sets are used, then the predictive distributions at a new location are generally different. Peruzzi [41] imposes sparsity post hoc via the graphical LASSO ([20]) after full spatial estimation of the cross-covariance in a two-stage plug-in approach. Krock et al. [30] directly model a sparse precision matrix for latent basis coefficients. In this approach, one does not model the covariance functions parametrically, making covariance properties like process smoothness and tail decay unavailable, and prediction at unobserved spatial locations is not considered.

One suitable framework for introducing sparsity is the notion of process-level conditional independence, as introduced by Brillinger [8] and Dahlhaus [13], which essentially amounts to sparsity in the inverse spectral density matrix. Approaches such as Dey et al. [17] attempt to model this sparsity from the spatial domain. In contrast, we develop a multivariate modeling approach in the spectral domain, directly targeting a sparse inverse spectral density, which results in a reference-free construction. Further, we use penalized likelihood inference to jointly estimate spatial dependence and sparsity in process dependence, avoiding a two-stage plug-in approach (e.g., Peruzzi, [41]). Moreover, our modeling approach separately parameterizes the multivariate dependence, modeled via a sparse graphical model, and the process-specific covariance parameters. Our approach also enables process-level conditional independence for processes with different range parameters, as well as processes with different classes of covariance functions, a noted gap in Peruzzi [41]. Inference proceeds via a random Fourier features approach [43, 18, 19, 38], and a computationally-efficient stochastic approximation expectation–maximization (SAEM) scheme [16] for penalized maximum likelihood inference is developed.

In simulations, we consider both when the processes are observed at the same locations (the colocated setting) and when there are at least some locations not observed for all processes (the misaligned setting). Asymptotic results suggest that in the colocated setting, leveraging additional processes for prediction of a single process may not yield improvements in prediction performance [54, 33]. However, many practical settings may have misaligned observations. While works in highly multivariate spatial processes have traditionally focused on the colocated setting [40, 17, 30], we demonstrate improved performance under misalignment. In addition, we examine quantities that may be of interest for the combined processes (e.g., the sum of the processes). For example, practitioners may be interested in aerosol column mass, obtained by summing the weighted mixing ratios of aerosol components across atmospheric layers, or the total toxic load, calculated as the sum of toxicity levels for different chemicals at each location. In this case, accurately capturing dependence between variables is necessary for appropriate uncertainty quantification of the predictions. We show improvements over existing approaches in terms of estimation, prediction accuracy, and predictive uncertainty quantification. We demonstrate this methodology in an application to the MERRA-2 aerosol data set [21]. We find substantial improvements in prediction for our approach in the misaligned setting. Such misalignment can arise when MERRA-2 fields are combined with observations at different locations or spatial resolutions [9, 55]. This demonstrates the benefits of modeling the underlying multivariate spatial process directly, rather than relying on independent univariate analyses.

Preliminaries and Prior Works

Let {Y∘(s)∈Rp,s∈Rd}\{Y^{\circ}(s) \in\mathbb{R}^{p}, s \in\mathbb{R}^{d}\} be a multivariate spatial process with output dimension pp and input dimension dd. We refer to the entries {Yj∘(s)∈R,s∈Rd}\{Y^{\circ}_{j}(s) \in\mathbb{R}, s \in\mathbb{R}^{d}\} of Y∘(s)Y^{\circ}(s) for j=1,…,pj = 1, \ldots, p as individual or marginal processes. A principal consideration for Gaussian processes is the covariance structure. In particular, if we assume E[Y∘(s)]=μ(s)\mathbb{E}[Y^{\circ}(s)] = \boldsymbol{\mu}(s) for some μ(s)∈Rp\boldsymbol{\mu}(s) \in\mathbb{R}^{p}, and the processes are second-order, the multivariate covariance function is:

E[(Y∘(s)−μ(s))(Y∘(t)−μ(t))T]=C(s,t;θ),\mathbb{E}\left[\left(Y^{\circ}(s)-\boldsymbol{\mu}(s)\right)\left(Y^{\circ}(t)-\boldsymbol{\mu}(t)\right)^{\mathsf{T}}\right] = C(s,t;\boldsymbol{\theta}),

depending on parameters θ\boldsymbol{\theta}. The Rp×p\mathbb{R}^{p \times p}-valued covariance function C(s,t;θ)C(s,t;\boldsymbol{\theta}) describes dependence both between and within processes at two locations. Let C(s,t;θ)=[Cjk(s,t;θ)]j,k=1pC(s,t;\boldsymbol{\theta}) = [C_{jk}(s,t;\boldsymbol{\theta})]_{j,k=1}^{p}, and we refer to Cjj(s,t;θ)C_{jj}(s,t;\boldsymbol{\theta}) as the marginal covariance function of process jj and Cjk(s,t;θ)C_{jk}(s,t;\boldsymbol{\theta}) as the cross-covariance function between processes jj and kk. We say that {Y∘(s)}\{Y^{\circ}(s)\} is second-order stationary (or henceforth, stationary) if E[Y∘(s)]=μ\mathbb{E}[Y^{\circ}(s)] = \boldsymbol{\mu} and C(s,t;θ)=C(t−s;θ)C(s,t;\boldsymbol{\theta}) = C(t-s;\boldsymbol{\theta}), so that the covariance function C(⋅,⋅;θ)C(\cdot,\cdot;\boldsymbol{\theta}) only depends on the lag between tt and ss through t−st-s [11]. Throughout, we assume stationarity and that μ=0p\boldsymbol{\mu} = \mathbf{0}_{p}.

In the stationary case, many multivariate covariances admit a spectral representation:

C(h;θ)=∫RdeihTxf(x;θ) dxC(h;\theta) = \int_{\mathbb{R}^{d}} e^{i h^{\mathsf{T}} x} f(x;\theta) \, dx

where f(x;θ)=[fjk(x;θ)]j,k=1p\mathbf{f}(x;\theta) = [f_{jk}(x;\theta)]_{j,k=1}^{p} is a matrix-valued spectral density and ii is the imaginary unit. If f(x;θ)\mathbf{f}(x;\theta) exists, the validity of the covariance C(h;θ)C(h;\theta) may be shown by ensuring f(x;θ)\mathbf{f}(x;\theta) is Hermitian and positive semidefinite for all x∈Rdx \in\mathbb{R}^{d}; see, for example, § 2.1 of Yarger et al. [53]. The spectral density representation is a special case of multivariate Bochner’s theorem, ensuring multivariate validity when spectral densities exist. In this paper, we present an approach that accommodates any stationary marginal covariance function with a spectral density, including mixed families of processes.

A key motivation for our work is the notion of process-level conditional independence ([13]). Let Y−jk∘(s)∈Rp−2\mathbf{Y}^{\circ}_{-jk}(s) \in\mathbb{R}^{p-2} be all entries of Y∘(s)\mathbf{Y}^{\circ}(s) except marginal processes jj and kk. Dahlhaus [13] defines process-level conditional independence for fixed j≠kj \ne k as Yj∘(s1)⊥ ⁣ ⁣ ⁣⊥Yk∘(s2)∣{Y−jk∘(s),s∈Rd}\mathbf{Y}^{\circ}_{j}(s_{1}) \perp\!\!\!\perp\mathbf{Y}^{\circ}_{k}(s_{2}) \mid\{\mathbf{Y}^{\circ}_{-jk}(s),s \in\mathbb{R}^{d}\} for all s1,s2∈Rds_{1},s_{2} \in\mathbb{R}^{d}, and shows that this process-level conditional independence holds whenever the j,kj,k entry of the inverse spectral density matrix f(x;θ)−1\mathbf{f}(x;\theta)^{-1} is 0 for all x∈Rdx \in\mathbb{R}^{d}. While conceptually appealing, one cannot practically condition on all {Y−jk∘(s),s∈Rd}\{\mathbf{Y}^{\circ}_{-jk}(s),s \in\mathbb{R}^{d}\} for a direct spatial domain construction. However, by directly modeling in the spectral domain, process-level conditional independence enables computational tractability when the conditional independence graph of the processes is sparse. The key bridge between our spectral construction to the spatial domain estimation and computation is provided by random Fourier features (see, e.g., Rahimi and Recht [43]).

Existing Approaches for Highly Multivariate Cross-covariances

Although already briefly summarized in § 1, we review relevant approaches for modeling highly multivariate spatial data with a bit more technical detail. First, Cressie and ZammitMangion (2016) construct a multivariate covariance one process at a time. Each subsequent variable depends on earlier variables across space, with a separate covariance for its remaining unexplained variation. The paper further uses a potentially-sparse directed acyclic graph to express dependence among variables. Alternatively, the stitching approach of Dey et al. [17] uses a finite set of locations S\mathcal{S} of size ∣S∣|\mathcal{S}|, referred to as reference or stitching locations, and a desired multivariate covariance function C(h;θ)C(h; \theta). They construct p2p^2 covariance matrices of size ∣S∣×∣S∣|\mathcal{S}| \times|\mathcal{S}|, denoted by Mjk(S,S)M_{jk}(\mathcal{S}, \mathcal{S}), for the reference locations such that conditional independence holds where desired for the spatial domain, and the covariances and cross-covariances match with Mjk(S,S)=Cjk(S,S)M_{jk}(\mathcal{S}, \mathcal{S}) = C_{jk}(\mathcal{S}, \mathcal{S}). The multivariate covariance is defined by:

Mjk(s1,s2)=Cjj(s1,S)Cjj(S,S)−1Mjk(S,S)Ckk(S,S)−1Ckk(S,s2).(1)M_{jk}(s_1, s_2) = C_{jj}(s_1, \mathcal{S}) C_{jj}(\mathcal{S}, \mathcal{S})^{-1} M_{jk}(\mathcal{S}, \mathcal{S}) C_{kk}(\mathcal{S}, \mathcal{S})^{-1} C_{kk}(\mathcal{S}, s_2). \tag*{(1)}

Essentially, the processes are stitched together at the reference locations S\mathcal{S}, then independently extended to all locations. This approach encourages sparsity in the inverse cross-spectral density matrix through a sparse underlying graph GG. However, the out-of-sample predictive variance computed in this manner at a new location sn+1∗s^*_{n+1} is dependent on the choice of reference set S\mathcal{S}. The sample locations and reference locations S\mathcal{S} may or may not be the same. Dey et al. [17] propose a chromatic Gibbs sampler for Bayesian estimation.

More recently, Peruzzi [40] proposed a similar approach with a few notable differences. Most importantly, they introduce an efficient likelihood estimation scheme that is computationally scalable to the setting where both nn and pp are large. The proposed model is similar to Dey et al. [17] in (1), using reference locations S\mathcal{S} with the choice of: Mjk(S,S)=LjLkTM_{jk}(\mathcal{S}, \mathcal{S}) = L_j L_k^{\mathsf{T}}, where LjL_j is the lower-triangular Cholesky decomposition of Cjj(S,S)C_{jj}(\mathcal{S}, \mathcal{S}). Therefore, the model has the same dependence on the reference location set S\mathcal{S} or its ordering. Peruzzi [41] examines spectral density structures that induce process-level conditional independence without a reference set. However, they note that their proposed framework does not accommodate general multivariate Matérn models with process-specific scale parameters. Whether these parameters can encode process-level conditional independence remains an open question that this paper addresses. For inference, Peruzzi [41] uses a two-stage plug-in approach.

Krock et al. [30] is conceptually the most similar to our work, in that they model directly in the frequency domain, obviating the need for a reference set. The p×pp \times p inverse spectral density matrix f(x;θ)−1\mathbf{f}(x;\boldsymbol{\theta})^{-1} is estimated directly and nonparametrically across xx using the graphical LASSO ([20]) and the fused graphical LASSO ([14]) penalties. This handles nonstationarities in highly multivariate data efficiently. On the other hand, process-level conditional independence is not specifically encouraged, in favor of piecewise conditional independence in the spectral domain.

Approximating Cross-covariances by Random Fourier Features

We outline the model construction and key ideas of our paper by illustrating the approximation of a multivariate cross-covariance following Emery et al. [19], where a full derivation is provided. We drop θ\boldsymbol{\theta} in the f(x;θ)\mathbf{f}(x;\boldsymbol{\theta}) notation for simplified development. Applying exp⁡(ihTx)=cos⁡(hTx)+isin⁡(hTx)\exp(\mathrm{i}h^{\mathsf{T}}x) = \cos(h^{\mathsf{T}}x) + \mathrm{i}\sin(h^{\mathsf{T}}x), the overall goal is to approximate the integral:

E[Y∘(s)Y∘(s+h)T]=∫Rd{cos⁡(hTx)ℜ[f(x)]−sin⁡(hTx)ℑ[f(x)]} dx,(2)\mathbb{E}\left[Y^{\circ}(s)Y^{\circ}(s+h)^{\mathsf{T}}\right] = \int_{\mathbb{R}^{d}} \left\{\cos(h^{\mathsf{T}}x)\Re[\mathbf{f}(x)] - \sin(h^{\mathsf{T}}x)\Im[\mathbf{f}(x)]\right\}\,\mathrm{d}x, \tag*{(2)}

using a set of random features in the spectral domain, where ℜ(z)\Re(z) and ℑ(z)\Im(z) are the real and imaginary parts of zz, respectively. First consider a proposal probability density function g(x)g(x) with g(x)>0g(x) > 0 wherever f\mathbf{f} is nonzero, and simulate Xℓ,j∼g(⋅)X_{\ell,j} \sim g(\cdot), ψℓ,j∼U[0,2π]\psi_{\ell,j} \sim U[0,2\pi] for j=1,…,pj = 1,\ldots,p and ℓ=1,…,L\ell= 1,\ldots,L independently. Here, LL indicates the size of the Fourier feature approximation and UU denotes the uniform density. Emery et al. [19] then construct:

Y∘(s)=1L∑ℓ=1L∑j=1paj(Xℓ,j)cos⁡(⟨Xℓ,j,s⟩+ψℓ,j)+bj(Xℓ,j)sin⁡(⟨Xℓ,j,s⟩+ψℓ,j).(3)Y^{\circ}(s) = \frac{1}{\sqrt{L}}\sum_{\ell=1}^{L}\sum_{j=1}^{p} a_j(X_{\ell,j})\cos\left(\langle X_{\ell,j},s\rangle+ \psi_{\ell,j}\right) + b_j(X_{\ell,j})\sin\left(\langle X_{\ell,j},s\rangle+ \psi_{\ell,j}\right). \tag*{(3)}

References

  1. [1]Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19(6):716–723.DOI
  2. [2]Allard, D., Clarotto, L., and Emery, X. (2022). Fully nonseparable Gneiting covariance functions for multivariate space-time data. Spatial Statistics, 52:100706.DOI
  3. [3]Apanasovich, T. V., Genton, M. G., and Sun, Y. (2012). A valid Matérn class of cross-covariance functions for multivariate random fields with any number of components. Journal of the American Statistical Association, 107(497):180–193.DOI
  4. [4]Argo (2023). Argo float data and metadata from Global Data Assembly Centre (Argo GDAC). doi:10.17882/42182.
  5. [5]Arroyo, D. and Emery, X. (2021). Algorithm 1013: An R implementation of a continuous spectral algorithm for simulating vector Gaussian random fields in Euclidean spaces. ACM Transactions on Mathematical Software, 47(1):1–25.DOI
  6. [6]Bolin, D. and Wallin, J. (2020). Multivariate type G Matérn stochastic partial differential equation random fields. Journal of the Royal Statistical Society Series B: Statistical Methodology, 82(1):215–239.DOI
  7. [7]Bond, T. C., Streets, D. G., Yarber, K. F., Nelson, S. M., Woo, J.-H., and Klimont, Z. (2004). A technology-based global inventory of black and organic carbon emissions from combustion. Journal of Geophysical Research: Atmospheres, 109(D14).DOI
  8. [8]Brillinger, D. R. (1996). Remarks concerning graphical models for time series and point processes. Brazilian Review of Econometrics, 16(1):1–23.DOI
  9. [9]Christensen, M. F., Heaton, M. J., Ruppert, S., Reese, C. S., and Christensen, W. F. (2019). Bayesian multi-scale spatio-temporal modeling of precipitation in the Indus watershed. Frontiers in Earth Science, 7:210.DOI
  10. [10]Clarke, F. H. (1990). Optimization and nonsmooth analysis. SIAM.
  11. [11]Cressie, N. (2015). Statistics for spatial data. John Wiley & Sons.
  12. [13]Dahlhaus, R. (2000). Graphical interaction models for multivariate time series. Metrika, 51:157–172.
  13. [14]Danaher, P., Wang, P., and Witten, D. M. (2014). The joint graphical lasso for inverse covariance estimation across multiple classes. Journal of the Royal Statistical Society Series B: Statistical Methodology, 76(2):373–397.pubmed.ncbi.nlm.nih.gov/24817823
  14. [15]Datta, A., Banerjee, S., Finley, A. O., and Gelfand, A. E. (2016). Hierarchical nearest-neighbor Gaussian process models for large geostatistical datasets. Journal of the American Statistical Association, 111(514):800–812.DOI
  15. [16]Delyon, B., Lavielle, M., and Moulines, E. (1999). Convergence of a stochastic approximation version of the EM algorithm. Annals of Statistics, 27(1):94–128.
  16. [17]Dey, D., Datta, A., and Banerjee, S. (2022). Graphical Gaussian process models for highly multivariate spatial data. Biometrika, 109(4):993–1014.DOI
  17. [18]Emery, X. and Arroyo, D. (2018). On a continuous spectral algorithm for simulating non-stationary Gaussian random fields. Stochastic Environmental Research and Risk Assessment, 32:905–919.DOI
  18. [19]Emery, X., Arroyo, D., and Porcu, E. (2016). An improved spectral turning-bands algorithm for simulating stationary vector Gaussian random fields. Stochastic Environmental Research and Risk Assessment, 30:1863–1873.DOI
  19. [20]Friedman, J., Hastie, T., and Tibshirani, R. (2008). Sparse inverse covariance estimation with the graphical lasso. Biostatistics, 9(3):432–441.pubmed.ncbi.nlm.nih.gov/18079126
  20. [21]Gelaro, R., McCarty, W., Suárez, M. J., Todling, R., Molod, A., Takacs, L., Randles, C. A., Darmenov, A., Bosilovich, M. G., Reichle, R., et al. (2017). The modern-era retrospective analysis for research and applications, version 2 (MERRA-2). Journal of Climate, 30(14):5419–5454.DOI
  21. [22]Genton, M. G. and Kleiber, W. (2015). Cross-covariance functions for multivariate geostatistics. Statistical Science, 30(2):147–163.DOI
  22. [23]Global Modeling and Assimilation Office [GMAO] (2015). MERRA-2 inst3_3d_aer_Nv: 3d, 3-hourly, instantaneous, model-level, assimilation, aerosol mixing ratio V5.12.4. Goddard Earth Sciences Data and Information Services Center (GES DISC), Greenbelt, MD, USA. Accessed: 2026-09-29.
  23. [24]Gneiting, T., Kleiber, W., and Schlather, M. (2010). Matérn cross-covariance functions for multivariate random fields. Journal of the American Statistical Association, 105(491):1167–1177.DOI
  24. [25]Goulard, M. and Voltz, M. (1992). Linear coregionalization model: tools for estimation and choice of cross-variogram matrix. Mathematical Geology, 24:269–286.DOI
  25. [26]Gui, K., Zhang, X., Che, H., Li, L., Zheng, Y., An, L., Miao, Y., Zhao, H., Dubovik, O., Holben, B., et al. (2026). Advancing operational global aerosol forecasting with machine learning. Nature, 651(8106):658–665.DOI
  26. [27]Guinness, J., Fuentes, M., Hesterberg, D., and Polizzotto, M. (2014). Multivariate spatial modeling of conditional dependence in microscale soil elemental composition data. Spatial Statistics, 9:93–108.DOI
  27. [28]Katzfuss, M. and Guinness, J. (2021). A general framework for Vecchia approximations of Gaussian processes. Statistical Science, 36(1):124–141.DOI
  28. [29]Kleiber, W. and Genton, M. G. (2013). Spatially varying cross-correlation coefficients in the presence of nugget effects. Biometrika, 100(1):213–220.DOI
  29. [30]Krock, M. L., Kleiber, W., Hammerling, D., and Becker, S. (2023). Modeling massive highly multivariate nonstationary spatial data with the basis graphical lasso. Journal of Computational and Graphical Statistics, 32(4):1472–1487.
  30. [31]Kuhn, E. and Lavielle, M. (2004). Coupling a stochastic approximation version of EM with an MCMC procedure. ESAIM: Probability and Statistics, 8:115–131.DOI
  31. [33]Lim, C. Y. and Wu, W.-Y. (2022). Conditions on which cokriging does not do better than kriging. Journal of Multivariate Analysis, 192:105084.DOI
  32. [35]Maring, H., Savoie, D. L., Izaguirre, M. A., Custals, L., and Reid, J. S. (2003). Mineral dust aerosol size distribution change during atmospheric transport. Journal of Geophysical Research: Atmospheres, 108(D19).DOI
  33. [36]Matérn, B. (1960). Spatial Variation: stochastic models and their application to some problems in forest surveys and other sampling investigations. Number 49:5 in Meddelanden från Statens skogsforskningsinstitut. Statens skogsforskningsinstitut.
  34. [37]Meng, X.-L. and Rubin, D. B. (1993). Maximum likelihood estimation via the ECM algorithm: A general framework. Biometrika, 80(2):267–278.DOI
  35. [38]Miller, M. J. and Reich, B. J. (2022). Bayesian spatial modeling using random Fourier frequencies. Spatial Statistics, 48:100598.DOI
  36. [39]Moreva, O. and Schlather, M. (2018). Fast and exact simulation of univariate and bivariate Gaussian random fields. Stat, 7(1):e188.DOI
  37. [40]Peruzzi, M. (2026a). Inside-out cross-covariance for spatial multivariate data. Journal of the American Statistical Association. doi:10.1080/01621459.2026.2640644.DOI
  38. [41]Peruzzi, M. (2026b). Partial correlation networks of Gaussian processes. arXiv preprint arXiv:2604.09953.arxiv.org/abs/2604.09953
  39. [42]Porcu, E., Alegria, A., and Furrer, R. (2018). Modeling temporally evolving and spatially globally dependent data. International Statistical Review, 86(2):344–377.arxiv.org/abs/1706.09233
  40. [43]Rahimi, A. and Recht, B. (2007). Random features for large-scale kernel machines. Advances in Neural Information Processing Systems, 20.
  41. [44]Randles, C. A., da Silva, A. M., Buchard, V., Colarco, P. R., Darmenov, A., Govindaraju, R., Smirnov, A., Holben, B., Ferrare, R., Hair, J., Shinozuka, Y., and Flynn, C. J. (2017). The MERRA-2 aerosol reanalysis, 1980 onward. Part I: System description and data assimilation evaluation. Journal of Climate, 30(17):6823–6850.DOI
  42. [45]Salvaña, M. L. O., Lenzi, A., and Genton, M. G. (2023). Spatio-temporal cross-covariance functions under the Lagrangian framework with multiple advections. Journal of the American Statistical Association, 118(544):2746–2761.
  43. [46]Schwarz, G. (1978). Estimating the dimension of a model. Annals of Statistics, 6(2):461–464.DOI
  44. [47]Speed, T. P. and Kiiveri, H. T. (1986). Gaussian Markov distributions over finite graphs. The Annals of Statistics, 14(1):138–150.DOI
  45. [48]Terres, M. A., Fuentes, M., Hesterberg, D., and Polizzotto, M. (2018). Bayesian spectral modeling for multivariate spatial distributions of elemental concentrations in soil. Bayesian Analysis, 13(1):1–28.
  46. [49]Ton, J.-F., Flaxman, S., Sejdinovic, D., and Bhatt, S. (2018). Spatial mapping with Gaussian processes and nonstationary Fourier features. Spatial Statistics, 28:59–78.DOI
  47. [50]Vecchia, A. V. (1988). Estimation and model identification for continuous spatial processes. Journal of the Royal Statistical Society Series B: Statistical Methodology, 50(2):297–312.DOI
  48. [51]Wei, G. C. G. and Tanner, M. A. (1990). A Monte Carlo implementation of the EM algorithm and the Poor Man’s data augmentation algorithms. Journal of the American Statistical Association, 85(411):699–704.
  49. [52]Yarger, D. and Bhadra, A. (2025). Multivariate confluent hypergeometric covariance functions with simultaneous flexibility over smoothness and tail decay. Mathematical Geosciences, 57(5):977–1001.arxiv.org/abs/2312.05682
  50. [53]Yarger, D., Stoev, S., and Hsing, T. (2026). Multivariate Matérn models – A spectral approach. Statistical Science, 41:69–98.arxiv.org/abs/2309.02584
  51. [54]Zhang, H. and Cai, W. (2015). When doesn’t cokriging outperform kriging? Statistical Science, 30(2):176–180.DOI
  52. [55]Zhang, J., Bonas, M., Bolster, D., Fuglstad, G.-A., and Castruccio, S. (2024). High-resolution global precipitation downscaling with latent Gaussian models and non-stationary stochastic partial differential equation structure. Journal of the Royal Statistical Society Series C: Applied Statistics, 73(1):65–81.DOI
  53. [56]Zhang, J., Crippa, P., Genton, M. G., and Castruccio, S. (2021). Assessing the reliability of wind power operations under a changing climate with a non-Gaussian bias correction. The Annals of Applied Statistics, 15(4):1831–1849.arxiv.org/abs/2011.01263

Paper details

vlnc:W_rn78r6mf
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents