Statistics Seminar

Date/Time:	Monday, 16 Nov 2015 from 4:10 pm to 5:00 pm
Location:	Snedecor 3105
Cost:	Free
URL:	www.stat.iastate.edu
Phone:	515-294-3440
Channel:	College of Liberal Arts and Sciences
Categories:	Lectures
Actions:	Download iCal/vCal | Email Reminder

"Bayesian Screening for Group Differences in High-Throughput Data", Eric Lock, Department of Biostatistics, University of Minnesota, Minneapolis

---

In modern biomedical research, it is common to screen for differences between sample groups in many variables that are measured using the same technology. Motivated by DNA methylation data, this talk focuses on screening for equality of group distributions for many variables with shared distributional features such as common support, common modes and common patterns of skewness. We propose a Bayesian nonparametric testing methodology, which improves performance by borrowing information across the different variables and groups through shared kernels and a hierarchical probability model for group differences. The inclusion of shared kernels in a finite mixture, with Dirichlet priors on the different weight vectors, leads to a simple framework for testing and we describe an implementation that scales well for high-dimensional data. We provide some theoretical results, including asymptotic rates of convergence, and consistency even under model misspecification. We compare with existing frequentist and Bayesian nonparametric testing methods, and describe an application to methylation data in brain tumors from the Cancer Genome Atlas.