Chasing tail(s): statistical intuition in high-throughput studies

Document Type

Presentation Abstract

Presentation Date

2-22-2010

Abstract

To overcome the inherent multiple-testing problem, many analyses of high-throughput data require extreme levels of significance, much further 'out in the tails' of standard reference distributions than usual. Other aspects of the analysis also require more care than usual. In the field of Genome-Wide Association Studies (GWAS) familiar practices have already been affected; for example the seminal Wellcome Trust GWAS showed the necessity of stringency in data-cleaning, and in control of confounding. These standards have been quickly adopted by the epidemiological community, but as GWAS moves forward, investigators are now beginning to attempt more subtle analyses, such as examining causal pathways, pleiotropy, and interactions. In this talk, we will discuss the challenges of getting appropriate regression tools to 'work' in a GWAS setting. For generic high-throughput settings, we also discuss the problem of interpreting multivariate regression results when attention is focused on only the most significant associations. For several problems, we show that 'rules of thumb' derived for traditional levels of significance can be unhelpful.

Additional Details

The speaker chairs the Analysis Committee of the Cohorts for Heart and Aging Research in Genetic Epidemiology (CHARGE) Consortium.

Monday, 22 February 2010
3:10 p.m. in Math 103
4:00 p.m. Refreshments in Math Lounge 109

This document is currently not available here.

Share

COinS