Biostatistics Collaborative Unit (BCU)

Printer-friendly versionPrinter-friendly version

Mission

The mission of the Biostatistics Collaborative Unit (BCU) is to enhance competitiveness of extramural grant proposals through improved study design and biostatistical methods and to promote collaborative research between biostatisticians and faculty in public health and biomedicine leading to the development of new and innovative statistical procedures.  The BCU accomplishes this through collaboration resulting in improved study design and appropriate application of state-of-the-art statistical methods to public health and biomedical data.  The BCU is a component of the Office of Clinical and Translational Research (OCTR) in the Office of Vice President for Research (OVPR). 

 

Consulting Services

Faculty are encouraged to seek statistical consultation when writing grants for extramural funding at least two months before the proposal deadline.  BCU services include post-award as well as pre-award biostatistical collaboration.  Services include:

  • Pre-award work.  Projects that require pre- and post-award collaboration or require development of innovative methods are handled through the BCU.  We collaborate on design, measurement, analysis of pilot data, methods of statistical analysis, power/sample size estimation and writing the statistical sections of applications and proposals.  These services are provided free of charge provided BCU faculty are involved as co-investigators on grant proposals. 
  • Post-award work.  Collaboration with investigators on the statistical analysis of data generated by the project includes data-base management, statistical analysis, original methodological research related to the analytical challenges posed by the project, and preparation of results for publication and presentation. 
  • Data Base Management.  Includes data entry, editing, maintaining data bases, checking for outliers or unusual observations.
  • Projects that do not require a sustained collaborative relationship or advanced statistical consulting will be generally handled by support staff such as the Research Associate-Biostatistician or graduate students.

 

BCU Staff Research Expertise

Biostatisticians of the BCU are trained in experimental and survey sampling design, and a variety of methods of biostatistical analyses including linear, generalized linear (e.g., logistic regression) and nonlinear models, hierarchical modeling, categorical data analysis, and survival analysis.  The following outlines the research interests of the BCU faculty. 

 

Stephen Rathbun (BCU Director and Professor)

Spatial Statistics.  Analysis of geo-referenced data; that is, data with spatial locations.  Data may include locations of specific events (e.g., residences of cancer cases), data on risk factors together with disease prevalence in regions (e.g., zip-codes, census tracts, counties), or data on concentrations of contaminants at selected sites.  Applications may be found in environmental epidemiology, health disparities, ecology, and sociology

Ecological Momentary Assessment.  Ecological Momentary Assessment uses electronic devices (e.g., smart phones) programmed to collect behavioral data in the every-day environments of study subjects.  Subjects may record salient events (e.g., cigarettes, dietary lapses, etc.) on the device, and the device may prompt subjects to answer a questionnaire at the times of the events, at random times or at specifically scheduled times (e.g., beginning or end of day).   Rathbun has conducted research on EMA study design and statistical analysis of resulting data including hierarchical modeling. 

 

Kevin Dobbin (Assistant Professor)

Cancer biomarkersare substances measured in bodily tissue or fluids that signify the presence or disposition of cancer, or related disease.  Biomarkers can be used to determine patient prognosis, diagnosis, and to identify the most effective treatment regime.  Dr. Dobbin has expertise in the development and validation of biomarkers and clinical assays, and their use in clinical trials.

High dimensional data occur when there are many variables on few subjects, and are produced by many modern biotechnological devices, such as next generation sequencing, RNA-Seq and microarrays.  Dr. Dobbin has extensive experience working with these types of data and has published widely on high dimensional study design and analysis. 

 

Hanwen Huang (Assistant Professor)

Statistical Genetics and Genetic Epidemiology.  Develop and evaluate statistical methods for the analysis of genetic data aimed at studying the role of genetic factors in determining health and disease in families and in populations, and the interplay of such genetic factors with environmental factors. Examples include the study of the aggregation of diseases in families, analysis of quantitative traits, genetic linkage studies and genetic association studies.  Huang has expertise in study design, data analysis, computational methods, and software development.

 

Ye Shen (Assistant Professor)

Infectious Disease Modeling.  Infectious diseases such as HIV, influenza, and schistosomiasis have different development and transmission patterns. Shen has experience with experimental design, data management, and analysis of a major HIV intervention study at the Chinese CDC, and a study examining telephone-administered psychotherapy for HIV-infected rural persons. Shen also has expertise in working with data for influenza infections and schistosomiasis control and elimination. 

Cancer Research Data Analysis.  In cancer research, repeatedly measured outcomes (e.g. CD4 counts) and time to event information (e.g. death/progression) are often of primary interest. Analyzing these data requires skills in longitudinal data modeling and survival analysis. Shen is familiar with both methodologies and is currently developing joint modeling approaches that combine the two types of information. Another research interest of Shen’s is on the analysis of data with missing values in oncology clinical trials.

 

Xiao Song (Associate Professor)

Survival Model with Measurement Error and Missing Data.  In survival analysis, a routine objective is to characterize the relationship between time to an event of interest (e.g., onset to AIDS) and some predictors. Standard inference procedures usually require observation of the true values of the covariates at the event times. However, some covariates cannot be accurately measured (e.g., CD4 count in HIV/AIDS research) or may be missing.  In addition, time-varying predictors are usually collected intermittently and may not be observed at the event times. Naive approaches that ignore measurement error and missing data might lead to biased estimation and erroneous inference.

Biomarker Data.  Biomarker data (e.g.  CD4 count, gene expression arrays) often arise in biomedical studies.  With multiple possible biomarkers, identification of important biomarkers can be made by variable selection.  The biomarkers can be used for medical diagnosis testing via ROC technology or treatment selection for patients in personalized medicine

 

Authorship Policy

Co-authorship on scientific articles is expected when substantive assistance is provided regarding study design and/or statistical analysis.

 

Access to Services

Consultation and services may be requested by contacting:

Stephen Rathbun, Director of BCU

rathbun@uga.edu

(706) 542-6302 office

 

Xinshuo Wang. Research Associate

xinshuow@uga.edu

(706) 713-2677 office