Search

We develop Bayesian predictive stacking for geostatistical models. Our approach builds an augmented Bayesian linear regression framework that subsumes the realizations of the spatial random field and delivers exact analytically tractable posterior inference conditional upon certain spatial process parameters. We subsequently combine such inference by stacking these individual models across the range of values of the hyper-parameters. We devise stacking of means and posterior densities in a manner that is computationally efficient without the need of iterative algorithms such as Markov chain Monte Carlo (MCMC) and can exploit the benefits of parallel computations. We offer novel theoretical insights into the resulting inference within an infill asymptotic paradigm and through empirical results showing that stacked inference is comparable to full sampling-based Bayesian inference at a significantly lower computational cost., Comment: 35 pages, 12 figures

A series of experiments in stationary and moving passenger rail cars were conducted to measure removal rates of particles in the size ranges of SARS-CoV-2 viral aerosols, and the air changes per hour provided by existing and modified air handling systems. Such methods for exposure assessments are customarily based on mechanistic models derived from physical laws of particle movement that are deterministic and do not account for measurement errors inherent in data collection. The resulting analysis compromises on reliably learning about mechanistic factors such as ventilation rates, aerosol generation rates and filtration efficiencies from field measurements. This manuscript develops a Bayesian state space modeling framework that synthesizes information from the mechanistic system as well as the field data. We derive a stochastic model from finite difference approximations of differential equations explaining particle concentrations. Our inferential framework trains the mechanistic system using the field measurements from the chamber experiments and delivers reliable estimates of the underlying physical process with fully model-based uncertainty quantification. Our application falls within the realm of Bayesian ``melding'' of mechanistic and statistical models and is of significant relevance to environmental hygienists and public health researchers working on assessing performance of aerosol removal rates for rail car fleets.

Multivariate functional or spatial data are commonly analysed using multivariate Gaussian processes (GP). Two recent approaches extend Gaussian graphical models (GGM) to multivariate Gaussian processes with each node representing an entire GP on a continuous domain and edges representing process-level conditional independence. The functional GGM uses a basis function expansion for a multivariate process and models the vector coefficients of the basis functions using GGM. Graphical GP directly constructs valid multivariate covariance functions by "stitching" univariate GPs, while exactly conforming to a given graphical model and retaining marginal properties. We bridge the two seemingly disconnected paradigms by proving that any partially separable functional GGM with a specific covariance selection model is a graphical GP. We also show that the finite term truncation of functional GGM is equivalent to stitching using a low-rank GP, which are known to oversmooth marginal distributions. The theoretical connections help design new algorithms for functional data that exactly conforms to a given inter-variable graph and also retains marginal distributions.

Multivariate spatially-oriented data sets are prevalent in the environmental and physical sciences.Scientists seek to jointly model multiple variables, each indexed by a spatial location, to capture anyunderlying spatial association for each variable and associations among the different dependent variables.Multivariate latent spatial process models have proved effective in driving statistical inference andrendering better predictive inference at arbitrary locations for the spatial process. High-dimensionalmultivariate spatial data, which is the theme of this article, refers to data sets where the number of spatiallocations and the number of spatially dependent variables is very large. The field has witnessed substantialdevelopments in scalable models for univariate spatial processes, but such methods for multivariate spatialprocesses, especially when the number of outcomes is moderately large, are limited in comparison. Here,we extend scalable modeling strategies for a single process to multivariate processes. We pursue Bayesianinference which is attractive for full uncertainty quantification of the latent spatial process. Our approachexploits distribution theory for the Matrix-Normal distribution, which we use to construct scalableversions of a hierarchical linear model of coregionalization (LMC) and spatial factor models that deliverinference over a high-dimensional parameter space including the latent spatial process. We illustrate thecomputational and inferential benefits of our algorithms over competing methods using simulation studiesand an analysis of a massive vegetation index dataset.

Joint modeling of spatially-oriented dependent variables are commonplace in the environmentalsciences, where scientists seek to estimate the relationships among a set of environmental outcomesaccounting for dependence among these outcomes and the spatial dependence for each outcome. Suchmodeling is now sought for very large data sets where the variables have been measured at a very largenumber of locations. Bayesian inference, while attractive for accommodating uncertainties through theirhierarchical structures, can become computationally onerous for modeling massive spatial data sets becauseof their reliance on iterative estimation algorithms. This manuscript develops a conjugate Bayesianframework for analyzing multivariate spatial data using analytically tractable posterior distributions thatdo not require iterative algorithms. We discuss differences between modeling the multivariate responseitself as a spatial process and that of modeling a latent process. We illustrate the computational andinferential benefits of these models using simulation studies and real data analyses for a Vege Indicesdataset with observed locations numbering in the millions.

The majority of Americans fail to achieve recommended levels of physical activity, which leads to numerous preventable health problems such as diabetes, hypertension, and heart diseases. This has generated substantial interest in monitoring human activity to gear interventions toward environmental features that may relate to higher physical activity. Wearable devices, such as wrist-worn sensors that monitor gross motor activity (actigraph units) continuously record the activity levels of a subject, producing massive amounts of high-resolution measurements. Analyzing actigraph data needs to account for spatial and temporal information on trajectories or paths traversed by subjects wearing such devices. Inferential objectives include estimating a subject's physical activity levels along a given trajectory; identifying trajectories that are more likely to produce higher levels of physical activity for a given subject; and predicting expected levels of physical activity in any proposed new trajectory for a given set of health attributes. Here, we devise a Bayesian hierarchical modeling framework for spatial-temporal actigraphy data to deliver fully model-based inference on trajectories while accounting for subject-level health attributes and spatial-temporal dependencies. We undertake a comprehensive analysis of an original dataset from the Physical Activity through Sustainable Transport Approaches in Los Angeles (PASTA-LA) study to ascertain spatial zones and trajectories exhibiting significantly higher levels of physical activity while accounting for various sources of heterogeneity.

Gaussian processes are widely employed as versatile modelling and predictive tools in spatial statistics, functional data analysis, computer modelling and diverse applications of machine learning. They have been widely studied over Euclidean spaces, where they are specified using covariance functions or covariograms for modelling complex dependencies. There is a growing literature on Gaussian processes over Riemannian manifolds in order to develop richer and more flexible inferential frameworks for non-Euclidean data. While numerical approximations through graph representations have been well studied for the Matérn covariogram and heat kernel, the behaviour of asymptotic inference on the parameters of the covariogram has received relatively scant attention. We focus on the asymptotic inference for Gaussian processes constructed over compact Riemannian manifolds. Building upon the recently introduced Matérn covariogram on a compact Riemannian manifold, we employ formal notions and conditions for the equivalence of two Matérn Gaussian random measures on compact manifolds to derive the parameter that is identifiable, also known as the microergodic parameter, and formally establish the consistency of the maximum likelihood estimate and the asymptotic optimality of the best linear unbiased predictor. The circle is studied as a specific example of compact Riemannian manifolds with numerical experiments to illustrate and corroborate the theory.

This paper describes and illustrates functionality of the spNNGP R package. The package provides a suite of spatial regression models for Gaussian and non-Gaussian point-referenced outcomes that are spatially indexed. The package implements several Markov chain Monte Carlo (MCMC) and MCMC-free Nearest Neighbor Gaussian Process (NNGP) models for inference about large spatial data. Non-Gaussian outcomes are modeled using a NNGP Polya-Gamma latent variable. OpenMP parallelization options are provided to take advantage of multiprocessor systems. Package features are illustrated using simulated and real data sets.

A key challenge in spatial statistics is the analysis for massive spatially-referenced data sets. Such analyses often proceed from Gaussian process specifications that can produce rich and robust inference, but involve dense covariance matrices that lack computationally exploitable structures. The matrix computations required for fitting such models involve floating point operations in cubic order of the number of spatial locations and dynamic memory storage in quadratic order. Recent developments in spatial statistics offer a variety of massively scalable approaches. Bayesian inference and hierarchical models, in particular, have gained popularity due to their richness and flexibility in accommodating spatial processes. Our current contribution is to provide computationally efficient exact algorithms for spatial interpolation of massive data sets using scalable spatial processes. We combine low-rank Gaussian processes with efficient sparse approximations. Following recent work by [1], we model the low-rank process using a Gaussian predictive process (GPP) and the residual process as a sparsity-inducing nearest-neighbor Gaussian process (NNGP). A key contribution here is to implement these models using exact conjugate Bayesian modeling to avoid expensive iterative algorithms. Through the simulation studies, we evaluate performance of the proposed approach and the robustness of our models, especially for long range prediction. We implement our approaches for remotely sensed light detection and ranging (LiDAR) data collected over the US Forest Service Tanana Inventory Unit (TIU) in a remote portion of Interior Alaska.

Gaussian processes (GPs) are widely used as distributions of random effects in linear mixed models, which are fit using the restricted likelihood or the closely-related Bayesian analysis. This article addresses two problems. First, we propose tools for understanding how data determine estimates in these models, using a spectral basis approximation to the GP under which the restricted likelihood is formally identical to the likelihood for a gamma-errors GLM with identity link. Second, to examine the data's support for a covariate and to understand how adding that covariate moves variation in the outcome y out of the GP and error parts of the fit, we apply a linear-model diagnostic, the added variable plot (AVP), both to the original observations and to projections of the data onto the spectral basis functions. The spectral- and observation-domain AVPs estimate the same coefficient for a covariate but emphasize low- and high-frequency data features respectively and thus highlight the covariate's effect on the GP and error parts of the fit respectively. The spectral approximation applies to data observed on a regular grid; for data observed at irregular locations, we propose smoothing the data to a grid before applying our methods. The methods are illustrated using the forest-biomass data of Finley et al.~(2008).

Internet usage in India has witnessed an enormous growth in last few years. The wireless data usage in 2017 has increased from 20,092 million GB per year from 828 million GB per year in 2014, showing a growth of more than 24 times. This increasing use of data is causing congestion in the existing bands which finally affects the quality of services provided to consumers. In order to cater to this surge in data consumption, it is essential that there should be a commensurate increase in the supply of broadband internet. Perhaps the approach to the supply also needs to change. Availability of additional spectrum is an important piece of this puzzle. V-band (57 GHz - 64 GHz) and E-band (71-76 GHz and 81-86 GHz) are two microwave bands which can be useful for bridging this need for additional spectrum. Spectrum in these bands can be used for high capacity data transmissions for last mile connectivity over short distances ranging from 200 meters to 3 km. These bands can be put to a variety of backhaul (i.e. for connecting towers). V-band can also be used for access under the Wi-Gig standards.

This note is devoted to the comparison between two Nearest-neighbor Gaussian processes (NNGP) based models: the response NNGP model and the latent NNGP model. We exhibit that the comparison based on the Kullback-Leibler divergence (KL-D) from the NNGP based models to their parent GP based model can result in reverse conclusions in different parameter spaces. And we suggest a heuristic explanation on the phenomenon that the latent NNGP model tends to outperform the response NNGP model in approximating their parent GP based model.

Species distribution models usually attempt to explain presence-absence or abundance of a species at a site in terms of the environmental features (so-called abiotic features) present at the site. Historically, such models have considered species individually. However, it is well-established that species interact to influence presence-absence and abundance (envisioned as biotic factors). As a result, there has been substantial recent interest in joint species distribution models with various types of response, e.g., presence-absence, continuous and ordinal data. Such models incorporate dependence between species response as a surrogate for interaction. The challenge we address here is how to accommodate such modeling in the context of a large number of species (e.g., order 10(2)) across sites numbering on the order of 10(2) or 10(3) when, in practice, only a few species are found at any observed site. Again, there is some recent literature to address this; we adopt a dimension reduction approach. The novel wrinkle we add here is spatial dependence. That is, we have a collection of sites over a relatively small spatial region so it is anticipated that species distribution at a given site would be similar to that at a nearby site. Specifically, we handle dimension reduction through Dirichlet processes, enabling clustering of species, joined with spatial dependence across sites through Gaussian processes. We use both simulated data and a plant communities dataset for the Cape Floristic Region (CFR) of South Africa to demonstrate our approach. The latter consists of presence-absence measurements for 639 tree species at 662 locations. Through both data examples we are able to demonstrate improved predictive performance using the foregoing specification.

In many applications, the dataset under investigation exhibits heterogeneous regimes that are more appropriately modeled using piece-wise linear models for each of the data segments separated by change-points. Although there have been much work on change point linear regression for the low dimensional case, high-dimensional change point regression is severely underdeveloped. Motivated by the analysis of Minnesota House Price Index data, we propose a fully Bayesian framework for fitting changing linear regression models in high-dimensional settings. Using segment-specific shrinkage and diffusion priors, we deliver full posterior inference for the change points and simultaneously obtain posterior probabilities of variable selection in each segment via an efficient Gibbs sampler. Additionally, our method can detect an unknown number of change points and accommodate different variable selection constraints like grouping or partial selection. We substantiate the accuracy of our method using simulation experiments for a wide range of scenarios. We apply our approach for a macro-economic analysis of Minnesota house price index data. The results strongly favor the change point model over a homogeneous (no change point) high-dimensional regression model.