STATISTICS COLLOQUIUM
Thursday, May12th, 2011 at 4.10pm, MSB 1147 (Colloquium Room)
Refreshments: 3.30pm, MSB 4110 (Statistics Lounge)
Speaker: Jiming Jiang (Dept. of Statistics, University of California, Davis)
Title: Selecting the Lasso/SCAD Regularization Parameters - The Adaptive Fence Ideas in Action
Abstract: The Lasso (Tibshirani, 1996) and SCAD (Fan and Li, 2001) are popular methods for regression model selection. These methods impose penalties on regression coefficients to shrink a subset of them towards zero to achieve the parameter estimation and model selection simultaneously. The amount of shrinkage is controlled by the regularization parameter. The procedures are consistent if the the regularization parameter satisfies some order conditions. The current methods of choosing the regularization parameter are cross-validation and various information criteria. In this work, a new data-driven method for choosing the regularization parameter is proposed and the consistency of the method is established. The method applies not only for the usual fixed-dimensional case but also for the divergent and NP-dimensionality situations. Simulation results show that the new method performs significantly better than the existing ones.
This work is joint with Zhen Pang and Bingqing Lin of the Nanyang Technological University, Singapore.
