The last decade has seen an explosion of interest in statistical modeling and analysis of spatiotemporally misaligned data and change-of-support problems, where different variables of scienti?c interest are observed at disparate scales making them dif?cult to be coherently modeled. This is especially relevant in environmental public health, where exposure data may be based upon data from monitoring data networks, while climate data are usually available as rasterized outputs from numerical models. The situation is further compounded by our objective of associating these factors with health outcomes (e.g. disease incidence, hospitalizations, mortality and so on), which are reported by public health sources as aggregated data over regions rather than at points. Furthermore, public health researchers today routinely encounter datasets exhibiting high-dimensional spatial misalignment or change-of-support, where ?dimension? refers to one or all of the following: (a) the number of spatial units (e.g., geographically referenced coordinates), (b) the number of temporal units (time points) at which the variables have been observed, and (c) the number of outcomes and other variables being studied. We propose a versatile collection of easily implementable and innovative Bayesian statistical methods that, in conjunction with appropriate software, will offer more comprehensive and statistically reliable mapping and analysis for misaligned spatiotemporal data in high-dimensional settings. Our methods and software will help spatial analysts to establish relationships among health outcomes and environmetal and climate-related predictors. Our dissemination efforts will deliver our methodology to a far broader audience of health and environmental researchers and administrators than is currently accessible.

Public Health Relevance

Identifying environmental and climate-related factors that are pronouncedly more detrimental to human health will improve the understanding and decision making process of health researchers, policy makers and patients, thereby having far-reaching bene?cial effects on the health care system and society. This will, how- ever, require statistical modeling and analysis for spatiotemporal databases on variables measured at highly disparate scales and exhibiting spatiotemporal misalignment. This application proposes to develop rigorous statistical methods based upon Bayesian principles and hierarchical models that will redeem investigators from using ad-hoc and qualitative methods that often produce biased estimates and can have far reaching bene?cial effects in public health research, perhaps touching unexpected corners of society.

National Institute of Health (NIH)
National Institute of Environmental Health Sciences (NIEHS)
Research Project (R01)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Joubert, Bonnie
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of California Los Angeles
Biostatistics & Other Math Sci
Schools of Public Health
Los Angeles
United States
Zip Code
Groth, Caroline; Banerjee, Sudipto; Ramachandran, Gurumurthy et al. (2018) Multivariate left-censored Bayesian model for predicting exposure using multiple chemical predictors. Environmetrics 29:
Bose, Maitreyee; Hodges, James S; Banerjee, Sudipto (2018) Toward a diagnostic toolkit for linear models with Gaussian-process distributed random effects. Biometrics 74:863-873
Gelfand, Alan E; Banerjee, Sudipto (2017) Bayesian Modeling and Analysis of Geostatistical Data. Annu Rev Stat Appl 4:245-266
Banerjee, Sudipto (2017) High-Dimensional Bayesian Geostatistics. Bayesian Anal 12:583-614