arXiv subjects

Christopher K. Wikle

Publications and source records attributed to Christopher K. Wikle.

58 records · Page 4Linked to original sources

arXiv2014

Spatial Fay-Herriot Models for Small Area Estimation with Functional Covariates

The Fay-Herriot (FH) model is widely used in small area estimation and uses auxiliary information to reduce estimation variance at undersampled locations. We extend the type of covariate information used in the FH model to include functional covariates, such as social-media search loads or remote-sensing images (e.g., in crop-yield surveys). The inclusion of these functional covariates is facilitated through a two-stage dimension-reduction approach that includes a Karhunen-Loève expansion followed by stochastic search variable selection. Additionally, the importance of modeling spatial autocorrelation has recently been recognized in the FH model; our model utilizes the intrinsic conditional autoregressive class of spatial models in addition to functional covariates. We demonstrate the effectiveness of our approach through simulation and analysis of data from the American Community Survey. We use Google Trends searches over time as functional covariates to analyze relative changes in rates of percent household Spanish-speaking in the eastern half of the United States.

stat.ME↗

arXiv2013

Modern Statistical Methods in Oceanography: A Hierarchical Perspective

Processes in ocean physics, air-sea interaction and ocean biogeochemistry span enormous ranges in spatial and temporal scales, that is, from molecular to planetary and from seconds to millennia. Identifying and implementing sustainable human practices depend critically on our understandings of key aspects of ocean physics and ecology within these scale ranges. The set of all ocean data is distorted such that three- and four-dimensional (i.e., time-dependent) in situ data are very sparse, while observations of surface and upper ocean properties from space-borne platforms have become abundant in the past few decades. Precisions in observations of all types vary as well. In the face of these challenges, the interface between Statistics and Oceanography has proven to be a fruitful area for research and the development of useful models. With the recognition of the key importance of identifying, quantifying and managing uncertainty in data and models of ocean processes, a hierarchical perspective has become increasingly productive. As examples, we review a heterogeneous mix of studies from our own work demonstrating Bayesian hierarchical model applications in ocean physics, air-sea interaction, ocean forecasting and ocean ecosystem models. This review is by no means exhaustive and we have endeavored to identify hierarchical modeling work reported by others across the broad range of ocean-related topics reported in the statistical literature. We conclude by noting relevant ocean-statistics problems on the immediate research horizon, and some technical challenges they pose, for example, in terms of nonlinearity, dimensionality and computing.

stat.ME↗

arXiv2013

Small Area Estimation via Multivariate Fay-Herriot Models with Latent Spatial Dependence

The Fay-Herriot model is a standard model for direct survey estimators in which the true quantity of interest, the superpopulation mean, is latent and its estimation is improved through the use of auxiliary covariates. In the context of small area estimation, these estimates can be further improved by borrowing strength across spatial region or by considering multiple outcomes simultaneously. We provide here two formulations to perform small area estimation with Fay-Herriot models that include both multivariate outcomes and latent spatial dependence. We consider two model formulations, one in which the outcome-by-space dependence structure is separable and one that accounts for the cross dependence through the use of a generalized multivariate conditional autoregressive (GMCAR) structure. The GMCAR model is shown in a state-level example to produce smaller mean square prediction errors, relative to equivalent census variables, than the separable model and the state-of-the-art multivariate model with unstructured dependence between outcomes and no spatial dependence. In addition, both the GMCAR and the separable models give smaller mean squared prediction error than the state-of-the-art model when conducting small area estimation on county level data from the American Community Survey.

stat.ME↗

arXiv2011

Semiparametric Bivariate Zero-Inflated Poisson Models with Application to Studies of Abundance for Multiple Species

Ecological studies involving counts of abundance, presence-absence or occupancy rates often produce data having a substantial proportion of zeros. Furthermore, these types of processes are typically multivariate and only adequately described by complex nonlinear relationships involving externally measured covariates. Ignoring these aspects of the data and implementing standard approaches can lead to models that fail to provide adequate scientific understanding of the underlying ecological processes, possibly resulting in a loss of inferential power. One method of dealing with data having excess zeros is to consider the class of univariate zero-inflated generalized linear models. However, this class of models fails to address the multivariate and nonlinear aspects associated with the data usually encountered in practice. Therefore, we propose a semiparametric bivariate zero-inflated Poisson model that takes into account both of these data attributes. The general modeling framework is hierarchical Bayes and is suitable for a broad range of applications. We demonstrate the effectiveness of our model through a motivating example on modeling catch per unit area for multiple species using data from the Missouri River benthic fish study, implemented by the United States Geological Survey.

stat.ME↗