Statistics Colloquium
November 1 (Fri) 11 a.m.
Alavi Commons Room, 6625 Everett Tower

Variable selection based on relevant information

Dr. Tanujit Dey
Department of Mathematics
College of William And Mary

In synchrony with the current advancement of science and technology, more complex and multi-dimensional data becomes ubiquitous. This status quo is both challenge and incentive for considering variable selection problems in generally ill-defined situations (number of variables in the data exceeding the number of observations). In this research, by assuming the framework of a linear model, our principle goal is to identify the relationship between certain outcomes and a few variables from the data set. Considering high to ultra-high dimensional cases, most variables are believed not responsible for expressing the outcome. Hence the objective would be to abandon those irrelevant variables from the data and hold on to only those that are more relevant for expressing the outcome. In order to do so, we propose a novel variable selection procedure along with supportive theoretical results. Extensive simulation studies have been performed to illustrate the performance of the proposed methodology in comparison with other familiar methods from the specialty literature. A number of real data sets have been analyzed from the point of view of the predictive performance to defend the novelty of the proposed methodology.

All statistics students are expected to attend.


Past colloquiums


Department of Statistics
3304 Everett Tower
Western Michigan University
Kalamazoo MI 49008-5152 USA
(269) 387-1420 | (269) 387-1419 Fax