High Dimensional Data, Covariance Matrices and Application to Genetics
Weill Medical College of Cornell University
Thursday, April 22, 2010, 12:30 pm in EMS E495A
In recent years there has been an explosion of high dimensional data which has challenged the conventional approach to statistics. This explosion is seen in a wide variety of fields ranging from genetics, to remote sensing, to financial engineering, and other fields. In this talk, I will focus on a special area of genetics called Quantitative Trait Loci mapping which deals with selecting possible genomic positions/loci (independent variables) associated with quantitative traits (dependent variables) in a linear regression setting. Typically, there are thousands of genetic loci to choose from and there are several (potentially thousands) of traits. I will propose a Bayesian framework to perform multivariate regression with model selection. In such a complex setting there are several issues, one of which relates to the covariance matrix and the effect of its prior on the Bayesian framework. I will present some investigation related to the prior of the covariance matrix in this context. I will also discuss an extension of this Bayesian framework to include gene-gene interaction, arbitrary covariates (fixed and random) and gene-environment interactions.
Refreshments will be provided in EMS E495B at 1:30.
All Colloquia and Seminars