题目：Recent Developments in Shrinkage Estimation for High-dimensional Data

报告人：Professor Tiejun TONG(香港浸会大学教授)

Abstract

High-dimensional data pose many challenges to traditional statistical and computational methods. Specifically, due to the small sample size, there are more uncertainties associated with standard estimations of parameters such as the mean and variance estimations. As a consequence, statistical analyses based on such parameter estimation are usually unreliable. To obtain more accurate parameter estimation some statistical methods, such as regularization through shrinkage, may yield better results.

In this talk, I will first review some new developments in shrinkage estimation on variances, under the framework of microarray data analysis. I will then introduce a recent work of mine that develops the shrinkage estimation of the population mean under the quadratic loss function for high-dimensional data. For illustration, a shrinkage-based diagonal discriminant rule will be proposed and it will be demonstrated to outperform the existing competitor in a wide range of settings.

Finally, I will finish the talk by mentioning some on-going projects, including the shrinkage estimation of covariance matrices project.

时间：2013年4月1日（周一）下午15:00开始

地点：数学系致远楼107室