Events and Seminars : 2013 Seminars

BIG DATA BIG BIAS SMALL SURPRISE

BIG DATA BIG BIAS SMALL SURPRISE

S. EJAZ AHMED, PH.D.
Professor and Dean
Department of Mathematics and Science
Brock University

TUESDAY, APRIL 15, 2014
2:00 p.m.– 3:00 p.m, CRB 692

In high-dimensional data settings where number of variables is greater than observations, or when number of variables are increasing with the sample size many penalized regularization approaches were studied for simultaneous variable selection and estimation. However, with the existence of covariates with weak effect, many existing variable selection methods may not distinguish covariates with weak signals and no signal. In this case, the prediction based on a selected submodel may not be highly efficient. In this talk, we propose a high-dimensional shrinkage estimation strategy to improve the prediction performance of a submodel. Such a high-dimensional shrinkage estimator (HDSE) is constructed by shrinking a ridge estimator in the direction of a predefined candidate submodel. Under an asymptotic distributional quadratic risk criterion, its prediction performance is explored. We show that the proposed HDSE performs better than the weighted ridge estimator. More importantly, it improves the prediction performance of any candidate submodel generated from most existing variable selection methods significantly. The relative performance of the proposed HDSE strategy is demonstrated by both simulation studies and the real data analysis.