Article
Physics, Multidisciplinary
Gareth W. Peters, Ido Nevat, Sai Ganesh Nagarajan, Tomoko Matsui
Summary: A class of non-Gaussian spatial random field models using Tukey g-and-h transformations for spatial field reconstruction was explored. The resulting warped spatial Gaussian process models support various desirable features and have wide applicability. Statistical properties of the models were carefully characterized to obtain flexible spatial field reconstructions, deriving five different estimators for important quantities in spatial field reconstruction problems. Simulation results and real data examples demonstrated the benefits of using Tukey g-and-h transformations over standard Gaussian spatial random fields in environmental monitoring applications.
Article
Statistics & Probability
Diego Morales-Navarrete, Moreno Bevilacqua, Christian Caamano-Carrillo, Luis M. Castro
Summary: Random fields are useful for representing complex dependence structures in natural phenomena. Gaussian random fields are commonly used due to their attractive properties. However, this assumption is restrictive for counting data. To address this, we propose a random field with a Poisson marginal distribution, generating a (non)-stationary random field that is mean square continuous and has Poisson marginal distributions. We provide analytic expressions for the covariance function and bivariate distribution of the proposed Poisson spatial random field. We investigate the weighted pairwise likelihood as a method for estimating the parameters in extensive simulations. Finally, we compare the proposed model with other models using real data.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2022)
Article
Agriculture, Multidisciplinary
Agda L. G. Oliveira, Joaquim P. Lima, Thiago L. Brasco, Lucas R. Amaral
Summary: This study evaluates the influence of considering anisotropy and trend in semivariogram modeling on the improvement of maps used in precision agriculture. The results indicate that modeling directional effects can improve the accuracy of kriging-generated maps. REML method performs better in strong anisotropy, while MoM method is more efficient in fields with weaker anisotropy.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2022)
Article
Statistics & Probability
Huiying Mao, Ryan Martin, Brian J. J. Reich
Summary: This article presents a new model-free nonparametric spatial prediction approach based on conformal prediction, which is applicable to complex spatial statistics problems. Numerical experiments demonstrate that the proposed prediction method has higher efficiency and validity for large datasets in various spatial settings.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2022)
Article
Geosciences, Multidisciplinary
Jianye Ching, Ikumasa Yoshida, Kok-Kwang Phoon
Summary: This paper compares two probabilistic models, sparse Bayesian learning (SBL) and Gaussian process regression (GPR), for the trend function of geotechnical spatial variability. The comparison is based on Bayesian evidence and shows that SBL usually outperforms GPR when the trend function can be represented by sparse basis functions (BFs), and vice versa. The paper also derives Kronecker-product formulae for the GPR method, which resolves the computational cost issue for 3D GPR analyses.
Article
Physics, Fluids & Plasmas
R. Erichsen Jr, A. Silveira, S. G. Magalhaes
Summary: The study investigates thermodynamic phase transitions of the joint presence of spin glass and random field, highlighting the importance of connectivity as a controllable parameter. Results show that for small connectivity, the RS solution remains stable at zero temperature, revealing significant differences with the predictions of the fully connected theory.
Article
Engineering, Civil
Mestapha Oumouni, Franck Schoefs
Summary: Non-destructive testing techniques are effective for assessing structures' condition and pathologies, but accurate diagnoses require numerous measurements with limited budget. Spatial variability of material properties can reduce the required measurements and help identify weak areas, but it still requires substantial measurements. A rational criterion incorporating accuracy and cost minimization is necessary to limit and optimize these measurements.
Article
Biology
Ben C. Stevenson, Rachel M. Fewster, Koustubh Sharma
Summary: Spatial capture-recapture models are commonly used to estimate animal density, but the assumption of independence in detection records at different detectors may not hold due to spatial correlation caused by animals moving around the survey region. This study introduces a latent detection field into the model to address this issue and highlights the predictable bias in SCR models with unmodeled spatial heterogeneity.
Article
Computer Science, Interdisciplinary Applications
Marco A. R. Ferreira, Erica M. Porter, Christopher T. Franck
Summary: Fast algorithms have been developed for Bayesian analysis of Gaussian hierarchical models with intrinsic conditional autoregressive (ICAR) spatial random effects. The algorithms, Spectral Gibbs Sampler (SGS) and Spectral Posterior Maximizer (SPM), are based on rewriting the hierarchical model in the spectral domain and proved the equivalence between two types of priors. These algorithms scale linearly with the sample size and provide computational advantages over existing algorithms in certain situations.
COMPUTATIONAL STATISTICS & DATA ANALYSIS
(2021)
Article
Engineering, Civil
Guotao Ma, Mohammad Rezania, Mohaddeseh Mousavi Nezhad
Summary: This paper proposes a stochastic method for analyzing the runout distance of sand collapse considering the spatial variability of shear strength. The method integrates random field theory and generalized interpolation material point method, and uses a Monte-Carlo simulation basis. The results show that deterministic analysis may underestimate the runout distance, while a heterogeneous model provides more realistic results. The uncertainty of the runout distance increases with the increase in the coefficient of variation of friction angle.
TRANSPORTATION GEOTECHNICS
(2022)
Article
Geosciences, Multidisciplinary
Roman Flury, Florian Gerber, Bernhard Schmid, Reinhard Furrer
Summary: This method identifies dominant features in spatial data by using multiresolution decomposition and variogram function estimation, capturing connected structures and patterns as well as assessing the width-extent of the dominant features.
SPATIAL STATISTICS
(2021)
Article
Economics
Ulrich K. Muller, Mark W. Watson
Summary: We propose a method for constructing confidence intervals that account for various forms of spatial correlation. The method controls coverage in finite sample Gaussian settings and performs well whenever the spatial correlation is weak.
Article
Geosciences, Multidisciplinary
Bethany J. Macdonald, Tilman M. Davies, Martin L. Hazelton
Summary: This article provides a comprehensive study of Monte Carlo maximum likelihood estimation (MCMLE) for log-Gaussian Cox processes (LGCPs). Various importance sampling algorithms for MCMLE are compared, and their performance is evaluated against other inference methods in numerical studies. The study finds that while the best MCMLE algorithm is practical for parameter estimation, its performance is sensitive to the choice of reference parameters defining the importance sampling distribution.
SPATIAL STATISTICS
(2023)
Review
Genetics & Heredity
Bader Arouisse, Tom P. J. M. Theeuwen, Fred A. van Eeuwijk, Willem Kruijer
Summary: The advances in high-throughput phenotyping have led to a greater number of secondary traits being observed, posing a challenge to improving genomic prediction for the target trait. Existing methods have limitations when dealing with a large number of secondary traits, emphasizing the need for novel approaches to enhance prediction accuracy.
FRONTIERS IN GENETICS
(2021)
Article
Economics
Guillaume Allaire Pouliot
Summary: This study presents methodology for regression analysis of misaligned data, where the independent and dependent variables do not coincide geographically. Two complementary methods are developed and investigated to avoid the need for covariance estimation or specification. A detailed reanalysis of Maccini and Yang (2009) reveals significant quantitative differences but largely sustains qualitative conclusions.
JOURNAL OF ECONOMETRICS
(2023)
Article
Statistics & Probability
Jiayi Wang, Raymond K. W. Wong, Xiaoke Zhang
Summary: In this article, a novel nonparametric covariance function estimation approach is proposed for multidimensional function data. The approach utilizes multilinear rank structures and reproducing kernel Hilbert spaces to model covariance operators and marginal structures flexibly. The resulting estimator is automatically semipositive definite and can incorporate various spectral regularizations, with trace-norm regularization promoting low ranks. The proposed approach achieves unified theoretical results for different sample sizes and data densities, revealing a phase-transition phenomenon from sparse to dense data.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2022)
Article
Computer Science, Artificial Intelligence
Jiyuan Tu, Weidong Liu, Xiaojun Mao
Summary: This paper presents a Byzantine-resilient method for distributed sparse M-estimation. By constructing a pseudo-response variable and transforming the optimization problem, a communication-efficient distributed algorithm is developed. Theoretically, it is proven that the proposed method achieves fast convergence and a support recovery result is established.
Article
Statistics & Probability
Rui Miao, Xiaoke Zhang, Raymond K. W. Wong
Summary: In this paper, the authors propose a two-step procedure to measure the dependency between two random functions using the Hilbert-Schmidt Independence Criterion (HSIC). They also introduce a new wavelet thresholding method for pre-smoothing and use Besov-norm-induced kernels for HSIC. The proposed method demonstrates superior numerical performance and interpretability in both simulation and magnetoencephalography (MEG) data applications.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Soil Science
Wenjuan Yu, Steven J. Hall, Haoyan Hu, Somak Dutta, Quanxin Miao, Jiaojiao Wang, Hongzhang Kang
Summary: Microbial responses to nitrogen enrichment under chronic ambient N deposition conditions are not well studied in subtropical forests. This study found that nitrogen enrichment may lead to increased phosphorus limitation in subtropical ecosystems, and the cooperation and competition between bacteria and fungi tend to weaken under nutrient-rich conditions. Additionally, ammonium and nitrate were significantly related to overall microbial community composition and other microbial groups involved in litter decomposition and N cycling.
SOIL BIOLOGY & BIOCHEMISTRY
(2022)
Article
Astronomy & Astrophysics
Mikhail M. Meskhi, Noah E. Wolfe, Zhenyu Dai, Carla Frohlich, Jonah M. Miller, Raymond K. W. Wong, Ricardo Vilalta
Summary: This study investigates the properties of the high-temperature nuclear equation of state (EOS), which impacts the type and mass of remnants after stellar collapse. By comparing synthetic populations with observed data, the study provides an evaluation of different EOS candidates.
ASTROPHYSICAL JOURNAL LETTERS
(2022)
Article
Plant Sciences
Henrique Uliana Trentin, Grigorii Batiru, Ursula Karoline Frei, Somak Dutta, Thomas Lubberstedt
Summary: The doubled haploid technology is a feasible and cost-efficient way to produce completely homozygous lines in maize, with factors such as haploid induction rate, donor background, and environmental conditions contributing to its success. Different genetic backgrounds showed varied haploid induction rates in different environments, with poor-performing donors possibly affected by anthocyanin inhibitor genes.
Article
Statistics & Probability
Xiaojun Mao, Zhonglei Wang, Shu Yang
Summary: This paper proposes a new matrix completion method for complex survey sampling, which allows for exploiting both row and column patterns simultaneously. The method is applied to assess the health status of the U.S. population.
ANNALS OF THE INSTITUTE OF STATISTICAL MATHEMATICS
(2023)
Article
Agronomy
Clayton N. Carley, Melinda J. Zubrod, Somak Dutta, Asheesh K. Singh
Summary: The symbiotic relationship between soybean roots and Bradyrhizobium japonicum bacteria leads to the development of nodules, which fix atmospheric nitrogen into ammonia for plant growth. The Soybean Nodule Acquisition Pipeline (SNAP) allows for easy quantification and evaluation of nodules in different soybean root system architectures. This study explores the unique relationships between nodules in taproot and non-taproot growth zones of diverse soybean genotypes, finding genotypic differences in nodule count, size, and total nodule area. The findings suggest a potential for enhanced nitrogen use efficiency and nodulation carbon to nitrogen production efficiency across soybean germplasm.
Article
Radiology, Nuclear Medicine & Medical Imaging
Subrata Pal, Somak Dutta, Ranjan Maitra
Summary: This article presents a method using deep learning techniques to enhance personalized synthetic MRI (syn-MRI) by implementing spatial regularization. The proposed method improves the quality of synthesized images even with limited training data.
MAGNETIC RESONANCE IN MEDICINE
(2023)
Article
Plant Sciences
Alper Adak, Myeongjong Kang, Steven L. Anderson, Seth C. Murray, Diego Jarquin, Raymond K. W. Wong, Matthias Katzfuss
Summary: High-throughput phenotyping (HTP) has not resulted in many new biological discoveries, but field-based HTP (FHTP) using UAVs has the potential to monitor plant population interactions with the environment. This study collected phenotypic data on maize lines in different environments and predicted complex traits using genomic and phenomic data. The study revealed a time-dependent association between genotypes and abiotic stresses, highlighting the importance of temporal phenomic data.
JOURNAL OF EXPERIMENTAL BOTANY
(2023)
Article
Mathematics, Interdisciplinary Applications
Wu Xue, Xiaoke Zhang, Kwun Chuen Gary Chan, Raymond K. W. Wong
Summary: This paper proposes a nonparametric covariate balancing method to estimate the counterfactual survival function by balancing covariates in a reproducing kernel Hilbert space (RKHS) using weights that are counterparts of inverse propensity scores. The method addresses the instability issue of the propensity score weighting method when there is limited covariate overlap between treatment and control groups. Simulation study and real data applications on smoking's causal effect on stroke patients' survival time and endotoxin's causal effect on female lung cancer patients' survival time demonstrate the appealing practical performance of the proposed method.
LIFETIME DATA ANALYSIS
(2023)
Article
Plant Sciences
Zihao Zheng, Bufei Guo, Somak Dutta, Vivekananda Roy, Huyu Liu, Patrick S. Schnable
Summary: Root lodging in maize is influenced by complex regulatory networks and genes associated with lodging resistance do not completely overlap with those related to root system architecture. The failure of root anchorage is a major cause of crop yield loss, but factors other than root system architecture may also play a role.
Article
Statistics & Probability
Shuoli Chen, Kejun He, Shiyuan He, Yang Ni, Raymond K. W. Wong
Summary: This article proposes a nonlinear Bayesian tensor additive regression model to predict tensor covariates with unknown shapes and discontinuous jumps. The proposed method uses a functional fused elastic net prior to model nonlinearity and spatial smoothness, detect discontinuous jumps, and identify active regions.
Article
Plant Sciences
Henrique Uliana Trentin, Recep Yavuz, Abil Dermail, Ursula Karoline Frei, Somak Dutta, Thomas Lubberstedt
Summary: The effectiveness of haploid induction systems depends on high haploid induction rate (HIR) and resource savings. An isolation field approach for hybrid induction is proposed. The success of efficient haploid production relies on inducer traits such as high HIR, abundant pollen production, and tall plants. Evaluation of seven hybrid inducers and their respective parents showed that mid-parent heterosis benefited plant height, ear height, and tassel size. Two promising hybrid inducers, BH201/LH82-Ped126 and BH201/LH82-Ped128, were identified for haploid induction in isolation fields. Hybrid inducers offer convenience and resource-effectiveness by improving plant vigor without compromising HIR.
Proceedings Paper
Computer Science, Artificial Intelligence
Yilang Zhang, Mengchu Xu, Xiaojun Mao, Jian Wang
Summary: This paper presents a novel method called CS-BGM that effectively expands the range of generator in generative compressed sensing. The method introduces uncertainties to the latent variable and parameters of the generator while adopting variational inference and maximum a posteriori to infer them. Theoretical analysis and extensive experiments demonstrate the improvement of CS-BGM over baselines.
INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 162
(2022)