报告题目：Tests for High Dimensional Data

报告人：祁永成 教授 （University of Minnesota Duluth，美国明尼苏达大学杜鲁斯分校）

时间：2019年7月5日 下午15：00-18：00

地点：数学馆 金融实验室

摘要：In classical multivariate analysis, statistical methods have been developed mainly for data from designed experiments, and dimensions of the data are fixed. Nowadays, new technology has generated various types of high dimensional data whose dimensions can be very large compared with the sample sizes.  In this case, some classical statistical approaches may not be applicable.  In this talk, we review some examples with particular interests in multivariate normal distributions and reveal the limitations in applying those classical statistical methods such as likelihood ratio test statistics to high dimensional data and when they can be extended to test on high dimensional data. We will also introduce some recent developments of non-parametric methods for ultra-high dimensional data.