Environmental Statistics Seminar

This seminar focuses on statistical issues related to assessing environmental effects on human health and analyzing environmental data in general. Specific areas of interest include air pollution epidemiology, exposure assessment, teratology, fertility and reproduction, respiratory studies, and community-based research as well as general topics such as errors-in-variables models, missing data methods, hierarchical modeling, smoothing, and methods for correlated data such as longitudinal and spatial data analysis. The seminars are generally pitched at a level that encourages student participation. Students interested in receiving credit for attending the seminars may sign up with individual faculty members for some guided readings on a special topic. Please see Chris Paciorek for details.

Chris Gennings, Ph.D.
Professor, Department of Biostatistics, Virginia Commonwealth University School of Medicine

"Analysis of Complex Chemical Mixtures: Integrating Component-based and Whole-Mixture Approaches"

ABSTRACT: Chemical mixture risk assessment methods fall into two general categories: component-based approaches and whole mixture approaches. Component-based approaches include construction of additivity models (e.g., default assumptions of dose addition or response addition) which have been used to test for departure from additivity at specified mixing ratios. Examples presented include the single chemical required method (SCR), the flexible single chemical required method (FSCR), and models for estimating interaction thresholds. Whole mixture approaches involve either direct evaluation of the mixture of concern or an assessment of the mixture of concern using data available on a “sufficiently similar” mixture. Combining the two general approaches permits tests for departure from additivity and estimation of benchmark doses at environmentally relevant mixing ratios with similar inference applied to sufficiently similar mixtures. The whole mixture strategy avoids default assumptions and is used to determine a boundary region for mixtures with similar inference to the reference mixture. The methods are illustrated with data from a study of a mixture of 18 polyhalogenated aromatic hydrocarbons (PHAHs) in rats exposed by oral gavage for four consecutive days. Serum total thyroxine (T4) was the response variable. Analysis of these data demonstrated a dose-dependent interaction among the 18 chemicals in the mixture, with additivity in the lower portion of the dose-response curve and synergy (greater than additive response) in the higher portion of the dose-response curve. Estimation of the interaction threshold within the observed experimental region suggested evidence of additivity in the low dose region. Total doses of the mixture that exceed the upper limit of the confidence interval on the interaction threshold were associated with a greater than additive interaction. The boundary of the similarity region for sufficiently similar mixtures to the experimentally observed reference mixture was used to determine similar mixture benchmark doses.

Brad Carlin, Ph.D.
Mayo Professor in Public Health, Division of Biostatistics, University of Minnesota School of Public Health

"Analysis of Marked Point Patterns with Spatial and Nonspatial Covariate Information"

ABSTRACT: Hierarchical modeling of spatial point process data has historically been plagued by computational difficulties. Likelihoods feature intractable integrals that are themselves nested within a Markov chain Monte Carlo (MCMC) algorithm. We extend customary spatial point pattern analysis in the context of a log-Gaussian Cox process model to accommodate spatially referenced covariates, individual-level risk factors, and individual-level covariates of interest that mark the process. We also use multivariate process realizations to capture dependence among the intensity surfaces across the marks. We illustrate using a collection of breast cancer case locations collected over the mostly rural northern part of the state of Minnesota that are marked by their treatment selection, mastectomy or breast conserving surgery ("lumpectomy"). The key substantive covariate (driving distance to the nearest radiation treatment facility) is spatially referenced, but other important covariates (notably age and stage) are not. Our approach facilitates mapping of marginal log-relative intensity surfaces for the two treatment options, and resolves the issue of whether women who face long driving distances are significantly more likely to opt for mastectomy while still accounting for all sources of spatial and nonspatial variability in the data. We also briefly discuss methods for statistical boundary analysis ("wombling") in such settings.

Roger D. Peng, Ph.D.
Assistant Professor, Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health

"On the Use of Spatial Information in Studies of the Acute Effects of Particulate Matter Constituents"

ABSTRACT: In air pollution epidemiology, there is an increasing focus on estimating the health effects of the chemical constituents of particulate matter (PM) and making inferences about health effects of sources of PM pollution. One issue that is raised by this new focus is how spatial properties of PM constituents effects subsequent health effects analyses and whether standard approaches can be applied to these data. Two issues that we will explore are spatial misalignment error in time series studies and the use of spatial information for source apportionment. Current approaches to time series modeling for estimating acute effects generally do not take into account the spatial properties of PM constituents and therefore could result in biased estimation of health risks. We present an approach for quantifying spatial misalignment error and show how adjusted heath risk estimates can be obtained using a plug-in approach and a two-stage Bayesian model. We apply our methods to a database containing information on hospital admissions, air pollution, and weather for 20 large urban counties in the United States.