The Department of Biostatistics is offering the following section of BIOS 691 (Special Topics in Biostatistics) in Spring 2010. Students are welcome to enroll.

BIOS 691 Applied Bayesian Biostatistics (Spring 2010)
Instructor: Dr. Nitai D Mukhopadhyay

Text: Bayesian Data analysis 2nd Edition, Gelman, Carlin, Stern, Rubin.

Software: Primarily R and Winbugs will be used.

Course content:

Introduction:

• Bayes rule, and basic probability theory.
• Likelihood principle.
• Definition of prior and posterior.
• Bayesian inference paradigm : defn of loss, posterior expected loss.
• Bayesian Estimation, credible sets and Bayesian model selection.

Univariate models:

• Conjugate priors, non-informative priors. Vague prior, improper prior.
• Reference prior. Jeffrey's prior.

Multiparameter models :

• Multivariate Normal
• Multinomial, Dirichlet process prior.

Hierarchical models,
Empirical Bayes methods.

Markov chain monte carlo simulation,

• Gibbs sampling,
• importance sampling.
• Metropolis Hastings method.
• Reversible Jump Markov chains .

Software: Winbugs, R introduction.

Model checking, convergence diagnostics, posterior predictive checking, sensitivity analysis.

Bayesian Applications:

• Bayesian linear models.
• Hierarchical Linear models.
• Generalized Linear models (Bayesian treatment),
• Non-linear models (examples).
• Large sample properties.

Bayesian model selection criterion and model averaging.

Time permitting we will cover Bayesian meta analysis, survival analysis and Bayesian Graphical models.

Computation: Winbugs and R will be used for the most part. If you are comfortable with C/C++ or Matlab, that will suffice as well. R and Winbugs will be introduced in class.

The Department of Biostatistics is offering the following section of BIOS 691 (Special Topics in Biostatistics) in Spring 2010. Students and researchers are welcome to enroll.

BIOS 691 Hierarchical Linear Models (Spring 2010)
Instructor: Dr. Yongyun Shin

Lecture: MW 4:00-5:20 pm, 3041 Theater Row

Prerequisite: a) one course introductory to hypothesis tests, probability distributions and estimation; b) one course in ANOVA and linear regression.

Applications: Multilevel hierarchical models have gained widespread popularity in education, social and behavioral sciences and public health applications recently. Application to contextual effects is pervasive in health and social and behavioral science studies. Numerous cluster-randomized trials have been conducted in epidemiology, health services research and treatment research studies with randomization of cities, housing developments, schools, classrooms, worksites, physicians and entire hospitals, rather than children or individual patients, to treatments while the outcome such as child academic achievement or patient response to the treatments varies at an individual level. More use of multilevel analysis for child health and academic achievement has been evidenced. A child attends a school nested within a neighborhood. The social setting also applies to patients of a doctor and doctors within a hospital, workers nested within a firm and firms within an industry, or adults living in a neighborhood and neighborhoods within a community where the longitudinal follow-up of individuals generates an additional level. Multilevel models may represent important sources of dependence associated with repeated measures within a child, children within a school, and schools within a neighborhood, ensuring appropriate uncertainty estimates such as standard errors and confidence intervals. These models can also represent heterogeneity of treatment effects across schools; the degree of such heterogeneity quantities the extent to which effects are context specific versus generalizable. Multilevel hierarchical models are appropriate to analyze these data, enabling researchers to address a wide range of seemingly unrelated research questions.

Course Material: This course introduces concepts, methods and applications in the analysis of multilevel hierarchical models. It will cover:
  1. review on a single-level regression model and ANOVA;
  2. extension of the single-level model to two levels and the notion of a level;
  3. impact of incorrectly specified number of levels on inferences;
  4. two-level hierarchical linear models and hierarchical generalized linear models
      a. intercept and slope as outcomes
      b. variance components and intra-class correlation
      c. contextual effects
      d. random coefficients;
  5. application of two-level hierarchical models
      a. cross-sectional
      b. longitudinal;
  6. estimation method for a two-level hierarchical linear model
      a. by maximum likelihood
      b. via the EM algorithm;
  7. extension to a three-level hierarchical linear model.
The course emphasis is on practical applications, not on theory. This course teaches students how to fit a hierarchical model via two software packages; SPSS and HLM 6. Knowledge of the software packages is not assumed. Students are welcome to bring their own data to analyze and write a paper/report for a group project. Students may also choose their own software packages for data analysis.

Jacob Wegelin (Biostatistics, Assistant Professor) presented a talk on the "Analysis Of Factors Affecting The Probability Of Finishing The Western States 100-mile Endurance Run" at the American College of Sports Medicine annual meeting in May 2009. He then presented "Analysis of longitudinal dichotomous data with non-nested grouping variables: a case study from sports science" at the Western North American Region of International Biometric Society in June 2009.

Mark Reimers (Biostatistics, Assistant Professor) and Phillip Yates (Biostatistics, Ph.D. Student) published a methodology article in BMC Bioinformatics titled "RCMAT: a regularized covariance matrix approach to testing gene sets." This methods paper resulted from their collaboration during the 2008 Summer Student Research Program.

Mark Reimers (Biostatistics, Assistant Professor) and Tobias Guennel (Biostatistics, Ph.D. Student) gave talks on microarray normalization at the NIH Microarray SIG seminar on September 2.

Faculty and students from the Department of Biostatistics attended the annual Joint Statistical Meetings in Washington, DC from August 1-6, 2009. Participants included:

• Kellie J. Archer (Biostatistics, Associate Professor) presented a poster on "Identifying Important Predictors Using L1 Penalized Models and Random Forests." Archer also co-authored a poster with Xiangrong Kong (Biostatistics, Ph.D. Alumnae 2008) titled "Variable Selection in Competing Risks Using the L1 Penalized Cox Model."
• Andre Williams (Biostatistics, Ph.D. Student) and Archer presented a poster, "Analysis of MicroRNA Data."
• Rhonda Ellis (Biostatistics, Ph.D. Student) served as a panelist for a "Success in Graduate School" discussion during the Pre-JSM Diversity Workshop "Fostering Diversity in Statistics."
• Tobias Guennel (Biostatistics, Ph.D. Student) presented a talk on "Noise Reduction For Array CGH Data Using Technical Covariates And Principal Component Analysis."
• Robert E. Johnson (Biostatistics, Associate Professor) attended the ENAR Regional Advisory Board meeting.
• Tina Cunningham (Biostatistics, Ph.D. Student) and Johnson presented a poster titled "Statistical Model and Sample Size Determination for the QuitNexus Study: Accounting for Within-Cluster Variability in the Denominator of the Counseling Rate."
• Nitai Mukhopadhyay (Biostatistics, Assistant Professor) gave a talk on "Testing and Credible Set Construction in High Dimension Using Multivariate Quantiles."
• Epiphanie Nyirabahizi (Biostatistics, Ph.D. Student) presented a talk on "Bayesian Approaches for Simultaneous Estimation of BMD with Multiple Chemicals and Multiple Endpoints."
• Mark Reimers (Biostatistics, Assistant Professor) presented a poster on "Expression array normalization using Principal Component Analysis."
• Roy Sabo (Biostatistics, Assistant Professor) gave a talk on "Hypothesis Testing for Various Dependence Structures."
• Yongyun Shin (Biostatistics, Assistant Professor) gave a talk on "Disparity in Causal Effects of Class Size on Academic Achievement Between Black and Other Students: Efficient Instrumental Variable Estimators with Tennessee Class Size Data Ignorably Missing," and also served as the session chair for a Section on Statistical Computing.

Mark Reimers (Biostatistics, Assistant Professor) spoke at the Science Museum of Virginia's Lunch Break Science program on August 12, 2009.

Karl Peace (Biostatistics, Ph.D. Alumnus 1976), has published his autobiography: Paid in Full.

The Department of Biostatistics held its annual Student Research Days on August 10-11, 2009 in the MCV Alumni House. See the program and abstracts for more information.

Phillip Yates (Biostatistics, Ph.D. Student), along with Nitai Mukhopadhyay (Biostatistics, Assistant Professor) and V. Ramakrishnan (Biostatistics, Associate Professor), presented a poster, titled "An inferential approach for one-sample gene network comparisons: assessing dissimilarity via a resampled local covering metric" at the Statistical Methods for the Analysis of Network Data in Practice Workshop held at University College Dublin from June 15-17, 2009.

The National Institute of Allergy and Infectious Diseases of the National Institutes of Health has awarded Virginia Commonwealth University the first year of a planned, four-year $11.5 million project to study how microorganisms found in the vagina influence health and disease in women. Mark Reimers (Biostatistics, Assistant Professor) is a co-principal investigator on the study. Read more details from the VCU News Service.

Tobias Guennel (Biostatistics, Ph.D. Student) presented "Noise Reduction For Array CGH Data Using Technical Covariates And Probe Level Information" on May 13, 2009 at the annual John C. Forbes Graduate Student Honors Colloquium.

Tina Cunningham (Biostatistics, Ph.D. Student) was selected to receive a School of Medicine Phi Kappa Phi Scholarship. Phi Kappa Phi is a national honor society promoting the pursuit of excellence in all academic fields.

The VCU Center on Human Needs recently launched a website for its flagship project, the Project on Societal Distress. The site is intended to provide the public and policymakers with accurate statistics about the prevalence of societal distress in five domains - hunger, housing, health, education, and income. Robert E. Johnson (Biostatistics, Associate Professor), Xiaoyan Deng (Biostatistics, Research Assistant), and Brian J. Bush (Biostatistics, Data Manager) continue to provide technical and statistical support for the project through BRADS. The Biostatistics Research and Data Services (BRADS) group provides data management and statistical analysis services to researchers in the VCU community.

Epiphanie Nyirabahizi (Biostatistics, Ph.D. Student) presented "Evaluating Neurotoxicity of a Mixture of five OP Pesticides Using a Composite Score" at the Society of Toxicology (SOT) annual meeting on March 16, 2009, in Baltimore, MD, and received an outstanding presentation award.