A Sequential Classification and Regression Tree Algorithm (S-CART) for Variable Selection in the Binary Classification Setting

Abstract:  The talk will discuss a sequential classification and regression tree algorithm (S-CART) in the context of variable selection in the binary classification setting. It will also briefly touch on the stochastic search variable selection (SSVS) technique in the context of multinomial classification. Results for S-CART, SSVS, and t-test application on real-world microarray datasets and datasets simulated from Bayesian models incorporating various degrees of data substructure and dependency will be discussed.

Refreshments and socializing: 9:30 - 10:00 AM  

Statistical Challenges in Vaccinomics: Host Genetic Factors and Immune Response to Anthrax Vaccine

Abstract:  Current methods for genetic association studies commonly involve straightforward phenotypic characterizations, such as case/control status. This presentation focuses on a more statistically challenging scenario motivated by an investigation into host genetic correlates of immune response to Anthrax Vaccine Adsorbed (AVA). The ethnically diverse study population was collected as part of a CDC sponsored clinical trial investigating the immunogenicity and reactogenicity of a reduced dose schedule and intramuscular injection of AVA. The primary outcome, antibody protective antigen (AbPA), measured longitudinally over a 43-month follow-up period, was subject to leftcensoring due to assay lower limits of quantification, and may reflect distinct stages of the immune response including components due to innate, inflammatory, and adaptive responses. A Bayesian hierarchical framework for the longitudinal profile of AbPA response to AVA is used to jointly account for these factors. Further, extensions employing multiple shrinkage, allowing for mode of inheritance uncertainty, are introduced through hierarchical Dirichet Process Mixture Models for simultaneously incorporating effects due to SNPs.

Refreshments and socializing: 2:30 - 3:00 PM