Statistical Science
Center for Statistical Research in the Environmental, Earth, and Life Sciences
The Center is a focal point for the research interests and accomplishments of several faculty members in the Department of Statistical Science. Most of these projects involve collaborations, both off-campus and on-campus, that are important to progress in the medical, environmental, and earth sciences. This research center fosters joint investigations of a rich variety of developments of new methodologies for data analysis in these particular areas of scientific research. Selected co-investigators and colloquium speakers enhance the intellectual pursuits of both faculty and graduate students with these interests.
Projects that are ideal for the Center draw upon a common set of statistical research specialties, which include
- modeling of multidimensional spatio-temporal dependence,
- statistical computing for diagnostic imaging
- nonparametric smoothing, and
- classification algorithms.
Some recent professional activities of Professors Gray, Gunst, Harris, Natarajan, Schucany and Woodward are described here. The list conveys the breadth of our commitment to these thriving branches of research within the discipline of statistical science.
LIFE SCIENCES
Functional Magnetic Resonance Imaging (fMRI) of Cognitive Functioning,
UTSWMC. Functional MRI of cognitive functioning is utilized increasingly for
the study of neurological disorders, including the evaluation of the response
to treatment. More needs to be known about the reproducibility of fMRI findings
for many cognitive tasks. A research team addressed this in a pilot study of
five healthy volunteers. Raji Natarajan is working with researchers at the University
of Texas Southwestern Medical Center (UTSWMC) on statistical inference for fMRI
data, with applications in the treatment of stroke
patients.
Dick Gunst, Wayne Woodward, and Bill Schucany are collaborating with medical imaging colleagues at the UTSWMC on a new initiative in functional magnetic resonance imaging (fMRI). The main initial thrust of the initiative is to secure funding to support statistical modeling of fMRI data that are to be taken on veterans of the Persian Gulf War who are suffering from what is now being called the Gulf War Syndrome. Dr. Robert Haley and colleagues at UTSWMC have characterized the Gulf War Syndrome using factor analyses of data collected from questionnaires administered to the veterans. Subsequent work has led the researchers to believe that fMRI might refine the understanding of the areas of the brain that have been affected. Current analyses of fMRI data do not comprehensively incorporate spatial and temporal variation in the measurements, two focal points of the research by the Gunst, Schucany, Woodward and graduate student research assistants from the Department of Statistical Science.
Meta-Analysis of a risk factor for Coronary Heart Disease, UTSWMC . The literature has several studies where low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), and plasma triglyceride levels have been associated with coronary heart disease (CHD). Raji and a collaborator are doing a meta-analysis, i.e., combine information from several studies, to assess the overall strength of association among these risk factors independently and to evaluate their joint effects upon the relative risk of CHD.
Administration of Amphetamine in the treatment of Aphasia, UTSWMC and TWU . A number of studies suggest that drugs that increase the release of norepinephrine promote recovery when administered late (days to weeks) after brain injury in animals. A small number of clinical studies have investigated the effects of the noradrenergic agonist, dextroamphetamine, in human patients recovering from motor deficits after they have had a stroke. In a pilot study it was determined that these findings extended to communication deficits subsequent to stroke, see Stroke (2001). Raji is working with collaborators at Texas WomanÕs University, UTSWMC, and Toronto on a larger trial to verify the results of the pilot study, which is under development.
Improving Learning of Technical Material by the Disabled . Buddy Gray has developed some software to help the handicapped learn statistics. His voice-activated technical word processor allowed a quadriplegic SMU undergraduate to successfully handle all of the material in our sophomore service course. This computer program also was essential to a blind masterÕs student completing his degree program that included a substantial amount of equations. Both NSF and NIH are considering how this package might be an essential assistance to people of all ages with disabilities.
New Methodology for Medical Data . How should one scatterplot data on twins? Mike Ernst, Rudy Guerra, and Bill Schucany answered this very basic question satisfactorily in a 1996 article in The American Statistician. A test of the broader class of bivariate interchangeability (which includes identical twins) was published in 1999 in the Journal of the American Statistical Association by Ernst and Schucany. Their results apply to medical baseline test/retest situations. Ian Harris has published a paper in Journal of Agricultural, Biological and Environmental Sciences in collaboration with Brent Burch and Roy St. Laurent of Northern Arizona University on the topic of measurement comparison. They investigate a type of correlation coefficient that is suitable for comparison of an approximate measurement with a more exact "gold standard" measurement. The method is applied to data on infarctions in canines.
Improved Diagnostic Tool for Heart Disease . Buddy Gray and Wayne Woodward are using signal-processing techniques in a fundamentally new research project using sonograms to classify patients with heart disease. This development promises to be a dramatic departure with greater accuracy than the present approach with electrocardiograms.
Other Biostatistical Projects . Patrick Carmack is doing basic research on classification trees for coronary heart disease.
ENVIRONMENTAL SCIENCES
Improving Chemical Modeling for Air Pollution Data . A multivariate receptor model is a latent variable model used to assess the composition and contribution of various pollution sources to an airshed. Dick Gunst is evaluating the effect of temporal dependence and measurement error correlation when using the effective variance solution to the chemical mass balance equations. Multivariate estimators that account for the multiple correlations in the data are under development.
Environmental Impact of Automobile Emissions. Automobile and truck
emissions are primary sources of atmospheric pollution that contribute to respiratory
and other health problems. The complexity of vehicle engine design and operation
and of combustion processes make the identification of ways to reduce emissions
an exceptionally difficult and expensive task. Dick Gunst has provided expertise
on the statistical design and analysis of experiments to automobile manufacturers
and major oil companies that contributed to major reductions in vehicle emissions.
Recent analyses of data from cooperative studies by both automobile manufacturers
and oil companies on the effects of fuel sulfur on emissions has led to tighter
EPA standards for permissible levels
of sulfur in fuels
New Data Analysis Tools for Ecology . Bill Schucany in collaboration with Pat Gerard, Mississippi State University, have developed some new techniques and published them in Biometrics and the Journal of Agricultural, Biological, and Environmental Statistics. They have extended the methodology for line-transect sampling in certain environmental settings. These flexible new estimators allow ecologists to more effectively combine such data on the same species of plants or animals from the field with fewer assumptions on the form of the detectability function away from the line. Sergio Juarez in dissertation research with Bill Schucany is evaluating new robust estimators for analyzing excedances over a threshold. The applications involve extremes in ozone measurements.
Spatial Modeling of Global Temperature Change . Temperature change
varies both over time and throughout spatial locations on the globe. Monthly
temperature measurements are available for over a century, but the locations
of the temperature measurements are concentrated in highly developed countries
and some well traveled ocean locations. The lack of a uniform pattern of environmental
monitoring stations throughout the globe presents formidable challenges to any
attempt to use station data to measure global climatic temperature change. Dick
Gunst has developed improved spatial models that accommodate the uneven spatial
density of monitoring stations around the globe. These models are contributing
to the calculation of more accurate estimates of global temperature change and,
equally important, the uncertainty in those
estimates.
Time Series Analysis of Global Temperatures . These data show a general increase over the past 100 years. There has been considerable controversy in the scientific community concerning whether this trend is a "deterministic" trend that should be predicted to continue or whether it is simply a "random" trending behavior that is part of the natural cycle of temperature changes. Through research funded by the Department of Energy, Wayne Woodward and Buddy Gray have developed methods for assessing whether an observed trend in a time series is deterministic or random.
EARTH SCIENCES
Seismic Signal Processing for Nuclear Monitoring. Wayne Woodward and
Buddy Gray have been involved with the "ultimate environmental protection
problem" of detecting underground nuclear explosions. Through funding from
the Department of Defense, they have developed new statistical outlier testing
methodology to be used with the data obtained from a global network of seismic
stations designed to detect unusual seismic events that may be linked to underground
nuclear testing.
Agricultural Sciences. In other research, Bill Schucany and Pat Gerard have devised a new technique for analyzing the freeze hardiness of plants. They produced a new method for identifying change points in thermal records. These discontinuities, minima, and inflection points can be valuable information for plant scientists investigating the cold tolerance of various species.
