
We propose an adjustment for the bias that occurs when an individual monitors a location and reports the status of an event. For example, a monitor may visit a plant each week and report whether the plant has flowered or not. The goal is to estimate the time the event occurred at that location. The problem is that popular estimators often incur bias because the event may not coincide with the arrival of the monitor and because the monitor may report the status in error. To correct for this bias, we use monotonic splines to estimate the event time. We first demonstrate the problem and our proposed solution. We then apply our method to a real-world example from phenology in which lilac are monitored by citizen scientists across the Northeastern USA, and the timing of the flowering is used to study anthropogenic warming. Our analysis suggests that current methods fail to account for monitoring bias and underestimate the peak bloom date of the lilac by 48 days on average. In addition, after adjusting for monitoring bias, several locations had anomalously late bloom dates that did not appear anomalous before adjustment. Supplementary materials accompanying this paper appear on-line.
This paper introduces a novel penalized spatial function-on-function quantile regression framework for analyzing spatially indexed functional data, bridging a critical gap between spatial functional models and quantile regression. Our work makes three key contributions. First, we propose the first spatial function-on-function quantile regression model that jointly accounts for spatial correlation across curves through a functional spatial autoregressive structure while allowing inference on arbitrary conditional quantiles of the functional response. Unlike traditional mean-based alternatives, this approach successfully captures state-dependent volatility and distributional dynamics beyond the conditional mean. Second, we develop a two-stage instrumental-variable estimation strategy to address endogeneity induced by the functional spatial lag. By utilizing tensor-product B-spline expansions with tensor-product roughness penalties, our method ensures optimal smoothness without the destructive information loss inherent in principal component truncation. Third, for fixed-spline dimensions, we establish √(n) -asymptotic normality of the spline coefficient estimators and the induced finite-rank Gaussian-process limits for the reconstructed coefficient surfaces. Extensive Monte Carlo experiments and a high-resolution analysis of Italian PM _2.5 air-quality data demonstrate that spatial function-on-function quantile regression significantly outperforms non-spatial and mean-based competitors, providing a robust and informative tool for environmental risk management and complex functional data analysis. Our method has been implemented in the package.
The increasing frequency and clustering of extreme environmental events, particularly wildfires, presents significant challenges for risk assessment and resource allocation. Traditional statistical models often fail to capture both the self-exciting nature of these events and their dependence on non-stationary climate drivers. We propose a novel Bayesian hierarchical framework that integrates climate covariates into a spatio-temporal Hawkes process model. Our approach decomposes wildfire risk into chronic (climate-forced) and acute (self-excited) components through a conditional intensity function where the background rate depends on slow-moving climate variables while the self-exciting component captures local clustering. We employ Hamiltonian Monte Carlo (HMC) for efficient posterior inference in this high-dimensional parameter space. Applied to historical wildfire data in the Western USA (1992-2023), our model demonstrates superior predictive performance compared to baseline Poisson and Cox process models. The framework provides interpretable risk decomposition and probabilistic forecasts under different climate scenarios, offering valuable insights for adaptive management and policy planning.
This study explores the prediction of a continuous target variable that is primarily measured as an ordinal outcome. Interest in such prediction may occur given data from double- or two-phase sampling designs, where an auxiliary variable is measured at all sampling units and a target and predictor variable, X, at a subset of those units. This study focuses on the case where X is nonnegative, is conditional on a zero process, and is sampled using a cluster sampling design. We model the ordinal outcome using cumulative logistic regression and the conditional regressor X, following log transformation, in a Bayesian setting. Estimation of model parameters and prediction of missing X is generally accurate and precise, particularly with sufficiently large Pr(X>0) and associated variances. This study is motivated by a need to accurately estimate the biomass of submersed aquatic vegetation on the Upper Mississippi River, where two types of measurements for biomass are available: inexpensive, ordinal biomass scores and expensive, continuous diver-harvested biomass data. Here, X represents plant biomass and Pr(X>0) is the usual species site occupancy measure. When fit using biomass data from the submerged aquatic vegetation species Vallisneria americana Michx, our model estimates parameters within the parameter space region where the model performed acceptably using synthetic data. We expect this model to appeal to investigators with ordered outcomes, cluster designs, and one or more continuous, positive predictors.
In the Earth system sciences, identifying physically meaningful patterns from dependent, high-dimensional, tensor-valued data is important for understanding variability in the Earth system. Orthonormal matrices are fundamental for statistical modeling of dependent data due to their role in matrix and tensor decomposition methods. While significant effort has been devoted to statistical inference for dependent matrix-valued data, these developments have not yet been extended to tensor-valued data. Here, we propose a general framework for dependent tensor-valued generalized linear mixed models (GLMMs) with a non-separable and non-stationary structured random effect. The random effect is parameterized using a tensor decomposition consisting of structured random orthonormal matrices, whose structures and associated coefficients are inferred during model estimation. Fully Bayesian inference enables joint estimation of all components of the dependent random effect and fixed effect within a unified statistical model. We validate our method with two synthetic examples and apply it to two real-world applications within atmospheric science, highlighting its capacity to uncover structured modes of variability across space, time, and vertical level. More broadly, our method provides a path forward for identifying structured patterns and their relative importance with Bayesian uncertainty quantification for arbitrary dependent, exponential family-distributed tensor-valued data.