Brian J. Reich

Adding Spatially-Correlated Errors Can Mess Up the Fixed Effect You Love

Many statisticians have had the experience of fitting a linear model with uncorrelated errors, then adding a spatially-correlated error term (random effect) and finding that the estimates of the fixed-effect coefficients have changed substantially. We show that adding a spatially-correlated error term to a linear model is equivalent to adding a saturated collection of canonical regressors, the coefficients of which are shrunk toward zero, where the spatial map determines both the canonical regressors and the relative extent of the coefficients’ shrinkage. Adding a spatially-correlated error term can also be seen as inflating the error variances associated with specific contrasts of the data, where the spatial map determines the contrasts and the extent of error-variance inflation. We show how to avoid this spatial confounding by restricting the spatial random effect to the orthogonal complement (residual space) of the fixed effects, which we call restricted spatial regression. We consider five proposed interpretations of spatial confounding and draw implications about what, if anything, one should do about it. In doing so, we debunk the common belief that adding a spatially-correlated random effect adjusts fixed-effect estimates for spatially-structured missing covariates. This article has supplementary material online.

Bayesian Spatial Quantile Regression

Tropospheric ozone is one of the six criteria pollutants regulated by the United States Environmental Protection Agency under the Clean Air Act and has been linked with several adverse health effects, including mortality. Due to the strong dependence on weather conditions, ozone may be sensitive to climate change and there is great interest in studying the potential effect of climate change on ozone, and how this change may affect public health. In this paper we develop a Bayesian spatial model to predict ozone under different meteorological conditions, and use this model to study spatial and temporal trends and to forecast ozone concentrations under different climate scenarios. We develop a spatial quantile regression model that does not assume normality and allows the covariates to affect the entire conditional distribution, rather than just the mean. The conditional distribution is allowed to vary from site-to-site and is smoothed with a spatial prior. For extremely large datasets our model is computationally infeasible, and we develop an approximate method. We apply the approximate version of our model to summer ozone from 1997–2005 in the Eastern U.S., and use deterministic climate models to project ozone under future climate conditions. Our analysis suggests that holding all other factors fixed, an increase in daily average temperature will lead to the largest increase in ozone in the Industrial Midwest and Northeast.

Flexible Bayesian quantile regression for independent and clustered data

Quantile regression has emerged as a useful supplement to ordinary mean regression. Traditional frequentist quantile regression makes very minimal assumptions on the form of the error distribution and thus is able to accommodate nonnormal errors, which are common in many applications. However, inference for these models is challenging, particularly for clustered or censored data. A Bayesian approach enables exact inference and is well suited to incorporate clustered, missing, or censored data. In this paper, we propose a flexible Bayesian quantile regression model. We assume that the error distribution is an infinite mixture of Gaussian densities subject to a stochastic constraint that enables inference on the quantile of interest. This method outperforms the traditional frequentist method under a wide array of simulated data models. We extend the proposed approach to analyze clustered data. Here, we differentiate between and develop conditional and marginal models for clustered data. We apply our methods to analyze a multipatient apnea duration data set.

A multivariate semiparametric Bayesian spatial modeling framework for hurricane surface wind fields

Storm surge, the onshore rush of sea water caused by the high winds and low pressure associated with a hurricane, can compound the effects of inland flooding caused by rainfall, leading to loss of property and loss of life for residents of coastal areas. Numerical ocean models are essential for creating storm surge forecasts for coastal areas. These models are driven primarily by the surface wind forcings. Currently, the gridded wind fields used by ocean models are specified by deterministic formulas that are based on the central pressure and location of the storm center. While these equations incorporate important physical knowledge about the structure of hurricane surface wind fields, they cannot always capture the asymmetric and dynamic nature of a hurricane. A new Bayesian multivariate spatial statistical modeling framework is introduced combining data with physical knowledge about the wind fields to improve the estimation of the wind vectors. Many spatial models assume the data follow a Gaussian distribution. However, this may be overly-restrictive for wind fields data which often display erratic behavior, such as sudden changes in time or space. In this paper we develop a semiparametric multivariate spatial model for these data. Our model builds on the stick-breaking prior, which is frequently used in Bayesian modeling to capture uncertainty in the parametric form of an outcome. The stick-breaking prior is extended to the spatial setting by assigning each location a different, unknown distribution, and smoothing the distributions in space with a series of kernel functions. This semiparametric spatial model is shown to improve prediction compared to usual Bayesian Kriging methods for the wind field of Hurricane Ivan.

The ecology of microscopic life in household dust

We spend the majority of our lives indoors; yet, we currently lack a comprehensive understanding of how the microbial communities found in homes vary across broad geographical regions and what factors are most important in shaping the types of microorganisms found inside homes. Here, we investigated the fungal and bacterial communities found in settled dust collected from inside and outside approximately 1200 homes located across the continental US, homes that represent a broad range of home designs and span many climatic zones. Indoor and outdoor dust samples harboured distinct microbial communities, but these differences were larger for bacteria than for fungi with most indoor fungi originating outside the home. Indoor fungal communities and the distribution of potential allergens varied predictably across climate and geographical regions; where you live determines what fungi live with you inside your home. By contrast, bacterial communities in indoor dust were more strongly influenced by the number and types of occupants living in the homes. In particular, the female : male ratio and whether a house had pets had a significant influence on the types of bacteria found inside our homes highlighting that who you live with determines what bacteria are found inside your home.

A hierarchical max-stable spatial model for extreme precipitation

Extreme environmental phenomena such as major precipitation events manifestly exhibit spatial dependence. Max-stable processes are a class of asymptotically-justified models that are capable of representing spatial dependence among extreme values. While these models satisfy modeling requirements, they are limited in their utility because their corresponding joint likelihoods are unknown for more than a trivial number of spatial locations, preventing, in particular, Bayesian analyses. In this paper, we propose a new random effects model to account for spatial dependence. We show that our specification of the random effect distribution leads to a max-stable process that has the popular Gaussian extreme value process (GEVP) as a limiting case. The proposed model is used to analyze the yearly maximum precipitation from a regional climate model.

Time-to-Event Analysis of Fine Particle Air Pollution and Preterm Birth: Results From North Carolina, 2001–2005

Exposure to air pollution during pregnancy has been suggested to be a risk factor for preterm birth; however, epidemiologic evidence remains mixed and limited. The authors examined the association between ambient levels of particulate matter <2.5 μm in aerodynamic diameter (PM(2.5)) and the risk of preterm birth in North Carolina during the period 2001-2005. They estimated the risks of cumulative and lagged average exposures to PM(2.5) during pregnancy via a 2-stage discrete-time survival model. The authors also considered exposure metrics derived from 1) ambient concentrations measured by the Air Quality System (AQS) monitoring network and 2) concentrations predicted by statistically fusing AQS data with process-based numerical model output (the Statistically Fused Air and Deposition Surfaces (FSD) database). Using the AQS measurements, an interquartile-range (1.73 μg/m(3)) increase in cumulative PM(2.5) exposure was associated with a 6.8% (95% posterior interval: 0.5, 13.6) increase in the risk of preterm birth. Using the FSD-predicted levels and accounting for prediction error, the authors also found significant adverse associations between trimester 1, trimester 2, and cumulative PM(2.5) exposure and preterm birth. These findings suggest that exposure to ambient PM(2.5) during pregnancy is associated with increased risk of preterm birth, even in a region characterized by relatively good air quality.

Simultaneous Factor Selection and Collapsing Levels in ANOVA

When performing an analysis of variance, the investigator often has two main goals: to determine which of the factors have a significant effect on the response, and to detect differences among the levels of the significant factors. Level comparisons are done via a post-hoc analysis based on pairwise differences. This article proposes a novel constrained regression approach to simultaneously accomplish both goals via shrinkage within a single automated procedure. The form of this shrinkage has the ability to collapse levels within a factor by setting their effects to be equal, while also achieving factor selection by zeroing out entire factors. Using this approach also leads to the identification of a structure within each factor, as levels can be automatically collapsed to form groups. In contrast to the traditional pairwise comparison methods, these groups are necessarily nonoverlapping so that the results are interpretable in terms of distinct subsets of levels. The proposed procedure is shown to have the oracle property in that asymptotically it performs as well as if the exact structure were known beforehand. A simulation and real data examples show the strong performance of the method.

Estimation and Prediction in Spatial Models With Block Composite Likelihoods

This article develops a block composite likelihood for estimation and prediction in large spatial datasets. The composite likelihood (CL) is constructed from the joint densities of pairs of adjacent spatial blocks. This allows large datasets to be split into many smaller datasets, each of which can be evaluated separately, and combined through a simple summation. Estimates for unknown parameters are obtained by maximizing the block CL function. In addition, a new method for optimal spatial prediction under the block CL is presented. Asymptotic variances for both parameter estimates and predictions are computed using Godambe sandwich matrices. The approach considerably improves computational efficiency, and the composite structure obviates the need to load entire datasets into memory at once, completely avoiding memory limitations imposed by massive datasets. Moreover, computing time can be reduced even further by distributing the operations using parallel computing. A simulation study shows that CL estimates and predictions, as well as their corresponding asymptotic confidence intervals, are competitive with those based on the full likelihood. The procedure is demonstrated on one dataset from the mining industry and one dataset of satellite retrievals. The real-data examples show that the block composite results tend to outperform two competitors; the predictive process model and fixed-rank kriging. Supplementary materials for this article is available online on the journal web site.

Variable Selection in Bayesian Smoothing Spline ANOVA Models: Application to Deterministic Computer Codes

With many predictors, choosing an appropriate subset of the covariates is a crucial—and difficult—step in nonparametric regression. We propose a Bayesian nonparametric regression model for curve fitting and variable selection. We use the smoothing splines ANOVA framework to decompose the regression function into interpretable main effect and interaction functions, and use stochastic search variable selection through Markov chain Monte Carlo sampling to search for models that fit the data well. We also show that variable selection is highly sensitive to hyperparameter choice, and develop a technique for selecting hyperparameters that control the long-run false-positive rate. We use our method to build an emulator for a complex computer model for two-phase fluid flow.