Conference on Nonparametric Statistics and Statistical Learning; Spring 2010; Columbus, OH

Ohio State University Research Foundation -Do Not Use, Columbus OH

Investigators

Abstract

An international conference on Nonparametric Statistics and Statistical Learning is scheduled for May 19-22, 2010 at the Ohio State University. The conference is cosponsored by the American Statistical Association (ASA) Section on Nonparametric Statistics and the newly formed ASA Section on Statistical Learning & Data Mining. This is the first and very timely effort to explore the synergy of these two areas in a single conference. The conference features six plenary talks by internationally prominent researchers whose work relates closely to both fields. Sixteen invited breakout sessions, each with three talks, cover additional topics with potential interest to both fields. These include Robustness of Statistical Learning Methods, Implications of Data Reduction by Data Depth, Nonparametric Bayesian Methods and Model Selection, Rank Based Methods for High Dimensional Problems, Designs for Variable Screening, Ranked Set Sampling and the Collapse of Importance Sampling in Very Large Scale Systems. There are also eight contributed paper sessions and two contributed poster sessions where junior investigators and graduate students are expected to participate. Nonparametrics and Statistical Learning share key foundational structures. Both disciplines avoid unrealistic assumptions about underlying distributions or models in scientific studies; both allow for complex association among variables; and both address problems in data summarization, discovery, classification and prediction. The advent of powerful computers with accompanying massive data sets brings both disciplines to the forefront of statistical theory and practice. The goal of the proposed conference is to present some of the most important recent advances in these fields and to discuss future research directions. A major part of the conference focuses on bringing statistical research leaders together with students, postdoctoral fellows, and young academics in a stimulating environment. The funding from the NSF supports attendance of graduate students and junior researchers in American universities to present either a talk or a poster. The conference is expected to accelerate interactions and collaborations among researchers in the important areas of nonparametric statistics and statistical learning, and thereby lead to the development of new and more effective methods of modeling and inference. Statistical learning, with its roots in nonparametric statistics, pervades virtually all aspects of modern life. From automatic interpretation of hand-written numbers to sort mail by zip code, to identification of which genes to target and which chemical features might best be incorporated into new medicines, to real time analysis of satellite images in predicting hurricane paths, to anywhere else there is massive data with many sources of variation, statistical learning has become the main strategy for putting the data to good use. Additional applications include optimal allocations of credit, selection of product features to emphasize in different markets, and evaluation of concordant sources when determining the need to raise threat levels to Homeland Security. The conference features a session on national statistics with key speakers from the Bureau of Labor Statistics, the Census Bureau, and the defense industry, and Application sessions with speakers from the National Center for Atmospheric Research, the software industry, and other private sectors. Interactions among academic, governmental, and industrial statisticians in the field should strengthen both research and practice.