Workshop in Biostatistics - archive

DATE: September 24, 2015
TIME: 1:30 - 3:00 pm
LOCATION: Medical School Office Building, Rm x303
TITLE: Two Problems in Large Scale Biostatistical Inference
SPEAKER: Michael Newton
Professor of Statistics and of Biostatistics and Medical Informatics,
University of Wisconsin-Madison

Abstract:
I will review findings from two PhD dissertations defended earlier this year, by Nick Henderson and Zhishi Wang.

A. Putting things in order: rvalues for ranking: In high-dimensional inference, the precision with which individual parameters are estimated may vary greatly among parameters, thus complicating the task to rank order parameters. I present a framework for evaluating different ranking/selection schemes as well as an empirical Bayesian methodology showing theoretical and empirical advantages over available approaches.

See Henderson and Newton, 2015 http://arxiv.org/abs/1312.5776.

B. Gene-set analysis: many sets at a time: A common problem in statistical genomics starts with a list of genes identified in some genome-wide analysis, and asks what known functional properties are over-represented in this gene list. Most enrichment statistics score each functional category (set) using gene level data for genes in that set, but ignore the other functional data on those same genes. Set overlaps thus complicate analyses and inhibit a concise functional description of the input gene list. I present a model-based multi-set approach to this task, including some of the interesting theoretical and computational challenges.

See Newton and Wang, 2015 http://www.annualreviews.org/eprint/A7Se8wheXD5rTtmituTk/full/10.1146/annurev-statistics-010814-020335
or Wang et al, 2015 http://projecteuclid.org/euclid.aoas/1430226091.