Environmetrics最新文献

This paper introduces a novel unit-Lindley mixed-effects model (NULMM) within the generalized linear mixed model (GLMM) framework, designed for analyzing correlated response variables bounded within the unit interval. Parameter estimation was conducted via maximum likelihood, using Laplace approximation and adaptive Gaussian- Hermite quadrature (AGHQ). Simulation studies revealed that the Laplace approximation yielded biased estimates, while AGHQ with 5 or 11 quadrature points produced unbiased results. The proposed model was applied to rural electricity access data from South Asian countries, with covariates including time, log(GDP), log(Rural Population), and income level. Results show that time and log(GDP) are positively associated with rural electricity access, whereas log(Rural Population) has a negative association but is not statistically significant. Additionally, significant disparities were observed between low-income and upper-middle-income countries. Model comparisons demonstrated that NULMM provides a better fit to the data than the beta mixed model and the unit-Lindley (UL) mixed model.

The analysis of spatiotemporal data is fundamental across multiple scientific disciplines, particularly in assessing the behavior of climate effects over space and time. A key challenge in this area is effectively capturing recurring climate phenomena, such as El Niño/La Niña (ENSO) phases, which induce prolonged periods of similar weather patterns across affected regions. To address this, our study introduces a novel spatiotemporal regression model that explicitly incorporates block structures representing these recurring climate effects. These blocks accommodate ENSO phases and manage the within-block correlations and shared characteristics, enhancing the model's ability to capture the influence of such phenomena on precipitation variability. The model further integrates functional predictors of both fixed and random nature, along with spatial covariance modeled via the Matérn class, to accommodate complex spatial, temporal, and block-related structures. Motivated by a monthly precipitation dataset from meteorological stations in Goiás State, Brazil, spanning 21 years (1980–2001), our approach assigns spatial effects to individual stations, temporal effects to months, blocks to ENSO phases, and repeated measures to years within those blocks. The results from simulation studies demonstrate the model's robustness and effectiveness, providing deeper insight into how recurring climate effects like ENSO impact rainfall patterns. This framework represents a significant methodological advancement in spatiotemporal modeling, highlighting the importance of explicitly modeling and estimating the effects of recurrent climate phenomena through block structures.

Air pollution remains a major environmental risk factor that is often associated with adverse health outcomes. However, quantifying and evaluating its effects on human health is challenging due to the complex nature of exposure data. Recent technological advances have led to the collection of various indicators of air pollution at increasingly high spatial-temporal resolutions (e.g., daily averages of pollutant levels at spatial locations referenced by latitude-longitude). However, health outcomes are typically aggregated over several spatial-temporal coordinates (e.g., annual prevalence for a county) to comply with survey regulations. This article develops a Bayesian hierarchical model to analyze such spatially-temporally misaligned exposure and health outcome data. We develop Bayesian predictive stacking for spatially and temporally misaligned data to optimally combine inference from multiple predictive spatial-temporal models. Stacking allows us to avoid iterative estimation algorithms such as Markov chain Monte Carlo that struggle due to convergence issues inflicted by the presence of weakly identified parameters. We apply our proposed method to study the effects of ozone on asthma in the state of California.

Bounded count data are commonly encountered in environmental studies. This paper examines two environmental applications illustrating their relevance. The first investigates the effect of winter malnutrition on mule deer (Odocoileus hemionus) fawn mortality. The second application analyzes public perceptions of environmental issues using data from the Eurobarometer 95.1 survey (March–April 2021), which includes a question rating the perceived severity of climate change on a scale from 1 to 10. Together, these studies demonstrate the need for flexible bounded count models in environmental research. In this context, the binomial and beta-binomial (BB) models are widely used for bounded count data, with the BB model offering the advantage of accounting for overdispersion. However, atypical observations in real-world applications may hinder the performance of the BB model and lead to biased or misleading inferences. To address this limitation, we propose the contaminated beta-binomial (cBB) distribution (cBB-D), which introduces an additional BB component to accommodate atypical observations while preserving the mean and variance structure of the BB model. The cBB-D thus captures both overdispersion and contamination effects in bounded count data. To incorporate explanatory variables, we further develop the contaminated BB regression model (cBB-RM), in which none, some, or all cBB parameters may depend on covariates. The proposed models are applied to two environmental datasets, complemented by a sensitivity analysis on simulated data to assess the influence of atypical observations on parameter estimation. The methodology is implemented in the open-source cBB package for R, available at https://github.com/arnootto/cBB.

Classic Bayesian methods with complex environmental models are frequently infeasible due to an intractable likelihood. Simulation-based inference methods, such as neural posterior estimation, calculate posteriors without accessing a likelihood function by leveraging the fact that data can be quickly simulated from the model, but converge slowly and/or poorly in high-dimensional settings. In this paper, we suggest that imposing strict variational assumptions on the form of the posterior can often combat these computational issues. Posterior distributions of model parameters are efficiently obtained by assuming a parametric form for the posterior, parametrized by the machine learning model, which is trained with the simulated data as inputs and the associated parameters as outputs. We show theoretically that if the parametric family of the variational posterior is correct, our posteriors converge to the true posteriors in Kullback–Leibler divergence. We also provide tools to help us identify if our parametric assumption is close to the true posterior, and modeling options if that is not the case. Comprehensive simulation studies using environmental models highlight our method's robustness and versatility. An analysis of the Zika virus in Brazil provides a thorough case study.

Procrustes analysis (PA) is a widely used method for aligning and comparing two or more multivariate data matrices and has been applied across various fields. However, PA is based on the least sum of squares, which has long been recognized as highly sensitive to outliers. This study proposes two new resistant alternatives to PA, advances five previously developed resistant PA methods, and provides accompanying R code for their implementation. The performance of PA and seven resistant PA methods in matching datasets containing outliers was compared and evaluated through simulation studies. We simulated 180 scenarios, including 36 scenarios without outliers and 144 scenarios varying in error variability, proportion of outliers, magnitude of outliers, sample size, reflection, and correlation structure. Results showed that in the presence of outliers, PA performed poorly in identifying outliers and matching datasets across all scenarios, while resistant PA methods, especially our proposed Improved Least Trimmed Squares, demonstrated strong robustness and practical utility. Key factors affecting methods' performance included error variability, proportion of outliers, magnitude of outliers, and sample size. These resistant PA methods have broad applicability in fields such as ecology and evolution, where robust comparison of multivariate datasets is essential.

The measurement and modeling of migration patterns and potential drivers are often limited by the quality of spatial data at local levels. Our aim is to address these constraints by advantaging of consolidated techniques in geospatial regression and emerging machine-learning approaches to explore how selected contextual socio-economic and environmental factors are related to migration trends. We define a three-step empirical strategy. First, we use the Geographically-Weighted-Regression to model the spatial variation at local scales. Second, we adopt the Multiscale-Geographically-Weighted-Regression to focus on the spatial heterogeneity across geographical scales. Third, we adopt the Geographically-Weighted-Random-Forest-Regression to validate variations across multiple local models. To illustrate, we apply the proposed methodology to selected environmental factors and gridded estimates of net migration patterns in Ghana, between 1985 and 2014. By the comparison of results, we argue the models' complementary explaining the contribution of each method to depict the spatial variability of migration environmental factors.