Events and Seminars : Upcoming Seminars

FAST AND ROBUST ASSOCIATION FOR HIGH-THROUGHPUT DATA

YI-HUI ZHOU, PH.D.
Research Assistant Professor
Bioinformatics Research Center
Departments of Biological Sciences
North Carolina State University
TUESDAY, SEPTEMBER 1, 2015
11:00 a.m.–12:00 p.m, CRB 692

Many years after the sequencing of the human genome, basic statistical issues of multiple testing remain important for discovery-based and translational science. However, at the extreme testing thresholds required for many -omics platforms, standard testing approaches can have highly inflated false positive rates, leading to false discoveries. Another problem, not always recognized by practitioners, is that standard approaches to analyses of “pathways” can also lead to numerous false discoveries. We present an approximation to exact association tests of trend that is accurate and fast enough for standard use in high-throughput settings. The approach is shown to be equivalent under permutation to likelihood ratio tests for the most commonly used generalized linear models.I also describe similar approaches for pathway testing using approximations to quadratic form statistics. Finally, I will discuss new asymptotic results for robust principal component analysis, which are useful as covariates in high-dimensional data. Together, the approaches constitute a unified and coherent approach to testing individual features and pathways.