Abstract:Given the computational challenges involved in calculating the maximum likelihood estimates for large spatial datasets, there has been significant interest in the research community regarding approximation methods for estimation and subsequent predictions. However, prior studies examining the evaluation of these methods have primarily focused on scenarios where the data are observed on a regular grid or originate from a uniform distribution of locations. Nevertheless, non-uniformly distributed locations are commonplace in fields like meteorology and ecology. Examples include gridded data with missing observations acquired through remote sensing techniques. To assess the reliability and effectiveness of cutting-edge approximation methods, we have initiated a competition focused on estimation and prediction for large spatial datasets with non-uniformly distributed locations. Participants were invited to employ their preferred methods to generate corresponding confidence and prediction intervals for synthetic datasets of varying sizes and spatial configurations. This competition serves as a valuable opportunity to benchmark and compare different approaches in a controlled setting. We evaluated the submissions from 11 different research teams worldwide. In summary, the Vecchia approximation and the fractional SPDE methods were among the best performers for estimation and prediction. Furthermore, the nearest neighbors Gaussian process and the multi-resolution approximation exhibited excellent performance in predictive tasks. These findings provide valuable guidance for selecting the most appropriate approximation methods based on specific data characteristics. Supplementary materials accompanying this paper appear online.

Convergent estimators of variance of a spatial mean in the presence of missing observations

Local Linear M‐estimation in Non‐parametric Spatial Regression

Spatial local M-estimation under association

Estimation of Semiparametric Varying-Coefficient Spatial Autoregressive Models with Missing in the Dependent Variable

Spatial confounding under infill asymptotics

Variance Estimation for Semiparametric Regression Models by Local Averaging

The Estimators of Mean for Semiparametric Varying-Coefficient Models with Response Missing at Random

A Consistent Nonparametric Estimation Of Spatial Autocovariances

Spatial Process Approximations: Assessing Their Necessity

Convergence Rate of Wavelet Density Estimator with Data Missing Randomly when Covariables Are Present

The Third Competition on Spatial Statistics for Large Datasets

Minimax Optimal Estimation of High-Dimensional Sparse Covariance Matrices with Missing Data.

Convergence and asymptotic normality of variational Bayesian approximations for exponential family models with missing values

Parameter Estimation in Spatial Autoregressive Models with Missing Data and Measurement Errors

On Statistical Properties of A Veracity Scoring Method for Spatial Data

Estimation of Spatial Panel Data Models with Randomly Missing Data in the Dependent Variable

Near-Optimal Mean Estimation with Unknown, Heteroskedastic Variances

Nonparametric Density Estimation for Data Scattered on Irregular Spatial Domains: A Likelihood-Based Approach Using Bivariate Penalized Spline Smoothing

Asymptotic Properties of Quasi-Maximum Likelihood Estimators for Heterogeneous Spatial Autoregressive Models

Spatially Varying Coefficient Models: Testing for Spatial Heteroscedasticity and Reweighting Estimation of the Coefficients

The Mean Imputation Estimators of Generalized Varying-ciefficient Models with Missing Data